> ## Content Index > Fetch the complete content index at: https://www.nearmedia.co/llms.txt > Use this file to discover other available public pages before exploring further. # EP 280- Why Google Deletes 2% of Reviews, Siri Search, Google’s New In-Stock Check & Its Ad Revenue Miss - URL: https://www.nearmedia.co/ep-280-why-google-deletes-2-of-reviews-siri-search-googles-new-in-stock-check-its-ad-revenue-miss/ - Published: 2026-10-29T19:35:26.000Z - Updated: 2026-10-29T19:37:50.000Z - Description: Greg Sterling and Mike Blumenthal dig into a study of Google review removals and why review spikes get punished, Siri's new local search in iOS 27.2, Google's Let's Confirm Its in Stock inventory agent, and the earnings miss Google partly blamed on AI Mode. - Author: Near Media - Tags: Conversations Greg Sterling and Mike Blumenthal take a week without a guest (next week it's [Myles Anderson of BrightLocal](https://www.brightlocal.com/research/smb-marketing-report/?ref=nearmedia.co) on his small business marketing report) to work through four stories: a [Reviewflowz and Cartelis study](https://www.reviewflowz.com/research/disappearing-google-reviews?ref=nearmedia.co) of which Google reviews disappear and why, Siri's much-improved search in iOS 27.2, Google's AI agent that phones stores to check stock, and an Alphabet quarter in which Google admitted AI Mode earns less than the search page it is replacing. Local search is a data business Google must both police and monetize — so as it purges review spikes, calls stores for inventory and herds users into AI Mode, businesses win by feeding it steady, structured signals, while Google is learning that the AI answers it is pushing earn far less than the ad-stacked page they replace. ## Takeaways - **Just under 2% of new reviews vanish within a week.** Reviewflowz and Cartelis tracked nearly five million reviews on Google profiles in the US and Europe for two weeks in August. About 1% of businesses accounted for half of all removals, and most lost few or none. Spain had the highest removal share, the US second. Home services and healthcare were hit hardest, hospitality least, and businesses with four to ten locations were the most exposed. *(00:51–05:59)* [\[1\]](#notes) - **The study only sees Google's second pass.** Mike's caveat: reviews that trip Google's word list never go live, so they can't be counted. What the study catches is the delayed AI analysis that runs when Google's servers have spare capacity, plus periodic sweeps that pull 50 to 500 year-old reviews when a new algorithm ships. That's why the removals cluster in high-spam categories like home services, not high-volume ones like restaurants and hotels. *(05:59–08:20)* - **Spikes trigger removals; content and employee names mostly don't.** Review length, images and stock phrases like "highly recommended" had no measurable effect, and naming staff didn't raise removal rates despite Google's new policy. Velocity did: bursts above a business's normal cadence drew deletions. Profiles with 85% or more five-star reviews saw twice the removal rate, and first-time reviewers were likelier to vanish. Mike notes nobody knows whether the "norm" is the business's own baseline or a Bayesian average across similar businesses. *(08:20–09:18, 13:21–13:41)* [Google contribution policy](https://support.google.com/contributionpolicy/answer/7400114?ref=nearmedia.co) - **Answer five-star reviews within 48 hours.** In the study, a five-star review with an owner response inside two days was about half as likely to be removed. Greg's broader point is that responses matter more to customers than to Google. *(04:00–05:00, 09:18–10:20)* - **Once you hit your baseline, send customers to a brand search.** Mike's tactic for businesses already in their market's top tier: replace the direct review link (which Google tracks) with an instruction to search the business name and click "write a review." It smooths out spikes and generates branded searches at the same time. *(13:42–14:56)* - **The 30-day review posting penalty is spreading.** Mike is seeing more home-services businesses whose new reviews simply stop appearing. Google sends no email, and support will confirm the penalty but won't lift it early. His read from this study and recent accounts: persistent spikes above the business's usual pace cause it. While it runs, keep asking via brand search so at least the searches count. *(14:34–16:39)* - **Review targets are relative to your market, not a number.** Greg cited BrightLocal's finding that many small businesses have fewer than 20 reviews as too few. Mike pushed back: check what competitors in your category and market have, because consumers pick among the top two or three volumes. He sees algorithmic bumps around 10 and possibly 100 reviews, and beyond that it's conversion, not rank. *(10:20–12:20)* [\[2\]](#notes) - **Only about a third of searchers read review text.** Surveys claim nearly everyone reads reviews, but Greg says most people look at star ratings and counts. Roughly a third, depending on category, check the most recent reviews, the negative ones and the owner responses. Mike adds that recency is how they confirm the business hasn't gone stale. *(12:20–13:20)* - **Siri is now a private, default local search engine.** In iOS 27.2, Siri answers web and local questions, saves searches in the Siri app, and returns legitimate local picks faster than ChatGPT, though slower than Google. Mike's tactics: keep your listing current in Apple Business with fresh photos and offers, earn Apple ratings (simple thumbs up or down) and Yelp reviews, and try to get onto Apple's Guides. His own reason for switching is privacy: he'd rather move most of his informational queries to Siri than keep feeding profiles at Gemini, ChatGPT and Claude. Greg expects it to erode Google usage but not revenue, since the questions Siri handles well (scores, conversions, where a show streams) carry few ads, and Siri still lacks Google's inventory data. *(17:27–24:57)* [\[3\]](#notes) - **Let Google Call took 48 hours and mostly backfilled from Merchant Center.** Mike's test for engagement rings: Google tried about 20 stores, one answered the AI call, and eight to ten more appeared because their inventory was in Google Merchant Center. Greg was still waiting after 36 hours, and the flow doesn't tell users how long it will take. The tactic: upload even a small or custom inventory to Merchant Center (Shopify and similar platforms do it for you) so you get backfilled, plus extra product visibility in the organic results. *(27:28–36:03)* [Google's how-to](https://blog.google/products-and-platforms/products/shopping/how-to-agentic-calling-let-google-call/?ref=nearmedia.co) [\[4\]](#notes) - **Google admitted AI Mode earns less than search.** Alphabet's ad revenue came in about $1.4 billion under expectations, and Chief Business Officer Philipp Schindler attributed part of it to "the lower monetization we were seeing in AI Mode." Shares fell about 10%. Greg's read: this is why Google won't flip the results page to AI Mode in 2027 — it can't fit the ten-plus ads of a typical local results page into an AI answer without driving users away. *(38:20–41:43)* [CNBC](https://www.cnbc.com/2026/10/28/alphabet-q3-earnings-search-ai.html?ref=nearmedia.co) [\[5\]](#notes) - **ChatGPT is the vulnerable one, because it lacks distribution.** Google can fund AI out of cash flow and push it through Search, Android, Chrome and Gmail; Apple owns the iPhone default. ChatGPT has about 1.2 billion weekly users, by Greg's rough estimate 9–12% of Google's daily search volume, and its ads have passed a $1 billion run rate — Greg even saw a prompt inviting local businesses to advertise. Anthropic avoided the fight by focusing on business customers. *(41:43–47:30)* [\[6\]](#notes) ### Related Reading - [Why Google reviews disappear (Reviewflowz and Cartelis)](https://www.reviewflowz.com/research/disappearing-google-reviews?ref=nearmedia.co) — the full study of 4.9 million reviews discussed in the first segment. - [Google Maps businesses lose 1.95% of new reviews in 7 days, study finds](https://ppc.land/google-maps-businesses-lose-1-95-of-new-reviews-in-7-days-study-finds/?ref=nearmedia.co) — PPC Land's breakdown, including the methodology and the vendor's interest in its findings. - [Review Removals, SMB Misalignment, AEO a Mistake?, Local Product Inventory, ChatGPT Search](https://www.nearmedia.co/review-removals-smb-misalignment-aeo-a-mistake-local-product-inventory-chatgpt-search/) — Greg's newsletter write-up of the same stories. - [iOS 27: what's new and what's coming next (Macworld)](https://www.macworld.com/article/2986799/ios-27-features-compatiblity-siri-ai-updates.html?ref=nearmedia.co) — what changed in Siri with iOS 27 and 27.2. - [Ask Google to call local businesses for you](https://blog.google/products-and-platforms/products/shopping/how-to-agentic-calling-let-google-call/?ref=nearmedia.co) — Google's guide to the agentic calling feature Mike demoed. - [What investors need to know about Alphabet's earnings report (CNBC)](https://www.cnbc.com/2026/10/28/alphabet-q3-earnings-search-ai.html?ref=nearmedia.co) — the Q3 miss and Schindler's AI Mode comment. - [EP 279 — Beyond the Link: How Citation Labs Rebuilt Its Agency for AI Visibility (Part 2)](https://www.nearmedia.co/ep-279-beyond-the-link-how-citation-labs-rebuilt-its-agency-for-ai-visibility-part-2/) — last week's episode. - [EP 271 — The Path of Enshittification: Apple Ads, Google's Toll Booth, and Meta's $18B Escape Hatch](https://www.nearmedia.co/ep-271-the-path-of-enshittification-apple-ads-googles-toll-booth-and-metas-18b-escape-hatch/) — the earlier hosts-only episode on ads, not AI, eating the click. - [EP 236 — Google's Review Algorithms Are Tightening: What GMBapi's Data Shows](https://www.nearmedia.co/ep-236-p-1-googles-review-algorithms-are-tightening-heres-what-gmbapis-data-shows/) — earlier review-removal data from Europe and the US. - [EP 196 — The Evolution of Local Inventory Solutions with Michael Carini](https://www.nearmedia.co/ep196-2/) — background on why local inventory has been so hard to get online. ## Practitioner Notes - **Pace review requests for home-services clients.** Home services had the highest removal rate in the study, and it's the category where Mike is seeing the 30-day posting penalty. For Near Media's home-services GBP work, that argues for a steady weekly request cadence pegged to the market's top tier, and a switch to brand-search requests once a client reaches it, rather than campaigns that create bursts. - **The brand-search review request is also a brand signal.** It turns every review ask into a branded query, the same behavior the brand series ([EP 268](https://www.nearmedia.co/ep-268-ai-overviews-brand-search-local-seo-cyrus-shepard-on-earning-the-click-from-the-archives/)) and Dana DiTomaso ([EP 278](https://www.nearmedia.co/ep-278-track-what-you-can-let-go-of-the-rest-dana-ditomaso-on-attribution-after-ga4/)) treat as the most useful signal a local business can watch in Search Console. - **Inventory feeds now decide whether an AI agent can find you.** With Let Google Call backfilling from Merchant Center and Siri lacking inventory altogether, structured product data is a visibility lever for local retailers, not just an ads requirement. It ties directly to the consumer-journey research Near Media ran for [Lastmile Retail (EP 273)](https://www.nearmedia.co/ep-273-fragmented-discovery-googles-eur460m-dma-fine-and-apples-distribution-bet/). - **Google's own earnings back "it's the ads, not the AI."** Schindler's admission that AI Mode monetizes worse is a public data point for the argument in [EP 271](https://www.nearmedia.co/ep-271-the-path-of-enshittification-apple-ads-googles-toll-booth-and-metas-18b-escape-hatch/) and [EP 275](https://www.nearmedia.co/ep-275-rankings-up-calls-down-what-ai-visibility-is-really-worth-to-local-businesses-part-2/): the ad-heavy results page is what is compressing local organic clicks, and Google has every reason to keep it. ### Concepts - **Multi-stage review filter** — Mike's model of Google's review moderation: a word-list check before a review goes live, delayed AI analysis when servers have capacity, and periodic sweeps of older reviews when a new algorithm ships. - **Review spike** — a burst of reviews above a business's normal cadence, measured against either the business's own history or a norm for similar businesses (Google doesn't say which). - **Review posting penalty** — a roughly 30-day block in which a profile's new reviews don't appear, applied without notice and confirmed only by support. - **Brand-search review request** — asking customers to search the business by name and leave a review from the profile, instead of sending a direct review link. - **Backfilling** — Let Google Call filling its answer with stores whose Merchant Center inventory shows the item when the stores themselves don't pick up. - **Monetization gap** — the difference between what Google earns from a traditional results page and from an AI Mode answer. - **Distribution advantage** — owning the default placement (Siri on iPhone; Gemini across Search, Android and Chrome), which drives AI usage more than model quality does. - **Enshittification in real time** — Greg's phrase for the pressure on Google and OpenAI to pack ads into AI products far faster than they filled the classic results page. ### Pick your starting point: [00:00](https://www.youtube.com/watch?v=vG25-nBI9ao&ref=nearmedia.co) Intro [00:51](https://www.youtube.com/watch?v=vG25-nBI9ao&t=51s&ref=nearmedia.co) Google review-removal study [05:59](https://www.youtube.com/watch?v=vG25-nBI9ao&t=359s&ref=nearmedia.co) What the study can't see [08:20](https://www.youtube.com/watch?v=vG25-nBI9ao&t=500s&ref=nearmedia.co) Spikes, not employee names [10:20](https://www.youtube.com/watch?v=vG25-nBI9ao&t=620s&ref=nearmedia.co) How many reviews you need [13:21](https://www.youtube.com/watch?v=vG25-nBI9ao&t=801s&ref=nearmedia.co) Ask for reviews via brand search [14:34](https://www.youtube.com/watch?v=vG25-nBI9ao&t=874s&pp=0gcJCWMAwfN6Pr3D&ref=nearmedia.co) The 30-day review posting penalty [17:27](https://www.youtube.com/watch?v=vG25-nBI9ao&t=1047s&ref=nearmedia.co) Siri search in iOS 27.2 [20:54](https://www.youtube.com/watch?v=vG25-nBI9ao&t=1254s&ref=nearmedia.co) Siri local results and tactics [23:40](https://www.youtube.com/watch?v=vG25-nBI9ao&t=1420s&ref=nearmedia.co) Moving queries to Siri for privacy [27:28](https://www.youtube.com/watch?v=vG25-nBI9ao&t=1648s&ref=nearmedia.co) Google's Let's Confirm Its in Stock [31:32](https://www.youtube.com/watch?v=vG25-nBI9ao&t=1892s&ref=nearmedia.co) The 48-hour wait [34:12](https://www.youtube.com/watch?v=vG25-nBI9ao&t=2052s&ref=nearmedia.co) 20 calls, one answer [38:00](https://www.youtube.com/watch?v=vG25-nBI9ao&t=2280s&ref=nearmedia.co) Google's earnings miss and AI Mode [42:45](https://www.youtube.com/watch?v=vG25-nBI9ao&t=2565s&pp=0gcJCWMAwfN6Pr3D&ref=nearmedia.co) ChatGPT's distribution problem [45:35](https://www.youtube.com/watch?v=vG25-nBI9ao&t=2735s&ref=nearmedia.co) Ads in AI answers[ ](https://www.youtube.com/watch?v=VuGPUEN%5F-mc&t=10s&ref=nearmedia.co) ### Full Transcript --> # Near Memo EP 280 (confirm number) — Review Spikes, Siri Search and the Ads AI Mode Can't Replace *Greg Sterling and Mike Blumenthal. Recorded October 29, 2026.* > **Editor's note:** This transcript has been lightly cleaned for reading. Stutters, false starts and clear transcription errors were corrected, and short backchannel ("yeah," "right," "mm-hmm") was folded into the main speaker's turn. Wording is otherwise faithful to what was said, including speaker-side factual slips, which are listed at the end rather than corrected inline. --- **Greg (00:10)** Hey everybody, welcome back to the Near Media Podcast, this week with me and Mike Blumenthal. We're taking a break from our parade of exciting guests that we've had over the last few weeks— **Mike (00:22)** Distinguished guests. **Greg (00:23)** Distinguished, wise and insightful guests. And next week, I believe— **Mike (00:26)** Yes. Although I would add to listeners that if you are interested in being a guest, don't hesitate to reach out to Greg's email, [gsterling@nearmedia.co](mailto:gsterling@nearmedia.co). We would love to have guests. **Greg (00:39)** Yes, that's fine. We're interested if you have an interesting point of view or interesting data; we're interested in talking to you potentially. We're gonna have Myles Anderson from BrightLocal next week to talk about their very interesting SMB marketing study, which they've done a number of, that showed, among other things, that small businesses aren't really giving Google the kind of emphasis and attention that it really deserves in local. We'll get into that in more detail, and why or why not that may be. This week we are going to talk about a new report which looks at review removal — say that ten times fast — and some of the implications there. We're gonna be talking about improved Siri local, well, not just local search but search in general; Google's Let Google Call feature, which is also Ask for Me — it's kind of part of the same cluster of AI-generated calls to local businesses to get information; and then finally, Google's earnings yesterday. Their ad revenue was off by almost a billion and a half dollars, which is a very rare miss for them, and we'll talk about why that is and the interesting implications for whatever transition to a more AI-centric user interface is ongoing. So, all right. The first item today is based on a report from a company called Reviewflowz and Cartelis, which is a French company. Together they did a study of review removal patterns on Google in the US and Europe, and they looked at about five million reviews during a two-week period in August. That's a caveat to the whole study, because it was a relatively short period. But what they found was that a meaningful number — just under 2% in the aggregate — of reviews were deleted within a week of appearing. And they noticed a number of patterns. It wasn't entirely consistent. There were some key verticals where the activity was more pronounced, such as home services and healthcare; hospitality, on the other end, had one of the lowest review removal rates. And then there were different findings by country. The country that had the highest share of new reviews removed was Spain. The US was number two, then Germany, France, Italy and other countries. And it seemed to be the case that companies with between four and ten locations were the most vulnerable. One of the more interesting findings was that about 1% of businesses accounted for roughly half of all the review removals. For some businesses there was no impact, and some businesses lost a handful of reviews. But there was a group of businesses, sort of 1% of the total, that saw a lot of activity. The report also said that one-star reviews were removed more frequently, about three times the rate of five-star reviews. But interestingly, businesses that had a lot of five-star reviews, above an 85% threshold — meaning 85% or more of their reviews were five stars — were more susceptible to review extraction than companies where there were fewer five-star reviews, where that 85% threshold was not crossed. Also, in five-star review cases where there was an owner response within 48 hours of the appearance of the review, there was a lower likelihood that the review would be removed. There are two more points here. Newbies, people that didn't have a track record on Google — this is very much like what Yelp does — were more likely to disappear than people who had a long track record on Google in terms of review volume. And that makes a certain amount of sense. They looked also at the content: text length, images, employee mentions, use of stock phrases like "this business is highly recommended," which might be prompted by the owner or might be some AI-generated thing. Most of those had no measurable impact on whether or not a review was removed. Review velocity, review spikes and a crossing of certain thresholds versus what the normal cadence was, did. So primarily in Europe, but also in the US, if there were these review spikes, then it was more likely that a certain number of reviews would be removed. So that's a very quick and dirty summary. If you're a subscriber to the newsletter, you can click through and get the full report, and if you want to look at it yourself, it's on the Reviewflowz website. We'll put it in the show notes so people can download it if they want. So Mike, what is your take on all that? **Mike (05:59)** Well, the first thing is that this only looks at reviews that went live and then were removed, not reviews that were removed before they went live. Google has a multi-stage review process, and they also worry about server load. In this multi-stage process, the first stage is just language. They have a word list, and if you use words that are inappropriate, the review just never shows up. This is a quick, easy thing for them to do. So if there are swear words, or it's racist or whatever — if you live on Dick Road and the customer mentions it, that might cause the review not to show up. Those kinds of things. Seriously, I'm not kidding. There is a Dick Road in our area, by the way. **Greg (06:39)** \[Laughs\] **Mike (06:45)** So we're only looking at a subset of total review removals: those that get analyzed by the newest algorithm after the fact, because the servers can't always run it in real time. They're lazy servers; they only run when Google has enough bandwidth. So if there are heavy loads from other things, they may get delayed. And also, as Google develops new, more thorough removal algorithms, they will go back slowly through the whole corpus and remove reviews. So periodically you'll see in the forum where some business lost 50 or 60 or 80 or 100 or 500 reviews that had been there for a year, as Google circles back through locations when they come out with a new review algorithm. So what we're seeing in this two-week snippet is likely mostly reviews that the delayed AI analysis applied to. And as you pointed out, it's much more common in high-stress categories like home services, where there's much more spam, and much less common in bricks-and-mortar restaurants and hotels, where the volume is high and the cheating is low. The one thing that I thought was pretty clear, though, is the fact that reviews with business employee names were not specifically targeted. Now, whether that's because they only get targeted if there's a certain volume of them — maybe there's a spike and a lot of employee names, maybe it's some sort of combined algo — but it seems primarily to be volume-based, and, as you pointed out, against some norm of the business. Now, whether the norm is a Bayesian norm — in other words, how many reviews should a business of this size get across our database — or whether it's an absolute norm against this business, we don't really know how Google is calculating that. **Greg (08:58)** I wouldn't be surprised if they have averages for all these categories and locations, and that's kind of the way they're— **Mike (09:03)** And volume, right? In other words, they would need to know roughly the volume. Maybe it's based on brand searches or something. So we don't know if this tabulation of what a spike is is an absolute or a relative value. **Greg (09:18)** Yeah. Let's talk about what the tactical implications are. One of the things that struck me about this is that they said, well, if you had images in your reviews— **Mike (09:30)** Responded to reviews. **Greg (09:32)** Well, response to reviews did seem to matter in the five-star review context. But SEOs get distracted, I think, by how to manipulate Google, or how to circumvent this or that, or how to comply with this or that rule that Google is talking about. And really, the important thing about reviews — my reaction to this — is to think about the customer experience. Customers are the ones that are looking at the reviews. If your reviews can't get by Google, then of course nobody sees them. But things like owner responses and images are meaningful to people, and people are the ones making the buying decisions: I'm gonna come to your restaurant, I'm gonna use you to redo my electrical panel, or whatever it is. Businesses really have to think less about Google — think about compliance, but less about Google and playing the game that Google wants everybody to play — and really think about the end customer. Because in our research, as you've talked about many times, reviews are number one or number two in terms of helping people make a decision. One of the things that was in that BrightLocal report was that most of the businesses that participated in their survey had fewer than twenty reviews. And that's really just not enough. So you have to— **Mike (11:02)** Well, it's very category-dependent, though, right? When people ask what the need is, what you need to do is go search in your market, in your categories, in your searches, and see what competitors have. What we've seen is people say, "This guy's got a thousand, this one's got seven hundred, I'm gonna go with a thousand." But if everybody has twenty, then twenty's fine. **Greg (11:22)** Yes, it is. But as a practical matter these days, most businesses have more than that. I mean, it depends. If you're in some tiny little hamlet somewhere in some obscure location, maybe you can't— **Mike (11:38)** I disagree. To me, generalizing that number is the same problem as overemphasizing reviews: "I gotta have X number of reviews." It just isn't the case. X is relative. **Greg (11:48)** No, it has to be benchmarked against your category and competition, of course. **Mike (11:53)** And once you have that number, you then need to review what you're doing to maximize client and SEO success. And it isn't necessarily getting to a thousand or getting to two thousand reviews. **Greg (12:06)** Not mindlessly. Not just relentlessly. **Mike (12:08)** Well, there does seem to be an algorithm bump at ten and maybe at a hundred. But beyond that, it's just a matter of consumer conversions. They prefer the top two or three review volumes. **Greg (12:20)** You need a process and you need to respond, right? And only twenty-five to thirty — fill in the blank — percent of people actually take the time to read review text. The surveys all say ninety-five percent of people read reviews. They do not. They look at star ratings, they look at the review counts. There is a meaningful subset of people — a third, let's call it, depending on the category and the audience — that will go in, look at the most recent review, look at negative reviews, look at the owner responses. But you do need to engage with those reviews. Engagement may be a signal to Google that's beneficial to the business in terms of its visibility, but for the end customer, if there's a terrible review, the customer wants to see the business owner's response. **Mike (13:15)** And they want to see if reviews are recent. They want to make sure the business hasn't become stale over the last year or two, whatever. But in terms of this report, to me what it says is two things. One is that, at least for now, Google does not seem to be penalizing employee names in reviews to any significant degree, but it does seem to be penalizing these spikes— **Greg (13:39)** Which assumes manipulation. **Mike (13:42)** To deal with those spikes, particularly if you're in home services, you need to understand your base review volume, and it's not always about going full-on to get as many more reviews as you possibly can. If you're hitting your base numbers, I might even change strategies with clients: instead of the review link, which Google tracks, give them your brand search. Let's say you're in the top 10% in your market and you're hitting your weekly numbers to stay there. Maybe you should switch the review link you send to a brand link, so that users who want to leave you a review do a brand search and click on the review link. It would not only prevent these spikes but would also give you more brand searches at Google. So there may be a way to deal with this tactically that maximizes more broadly for the consumer experience and for your rank. The other thing related to this is that I have been seeing an increase in the number of businesses that are receiving the review posting penalty. Here's the situation: a customer came to me and said, "We haven't had a review in two weeks." I looked at their account and said, "You know, it's because you're getting so many reviews and you're in a home service industry, a high-spam category, that Google gave you a penalty." And they said, "But they didn't send us an email. Wouldn't they send us an email?" I said, "Usually not. You've gotta reach out to support. Support will reply back that you're under a penalty, and you've got to suck it up." And they said, "But isn't there anything we can do? We're trying to do this honestly," blah, blah, blah. And it's like, no, there isn't. You've got this penalty. It's likely to last 30 days. The question is, during those 30 days, what should you be doing? Should you be going for your average rate? Should you not be doing anything? My suggestion to people who've had this penalty is to switch to the brand search idea for asking for reviews. Send your customers out there. However many leave reviews, leave reviews; if they don't, they don't, but at least you're getting a brand search out of it. And hopefully at the end of 30 days Google will lift the penalty. I would reach out to support immediately if you get this penalty. What I think causes it, upon reading this research and looking at the couple of accounts I've looked at over the past week or two, is persistent spikes. In other words, you've got a level that you're going at, or that Google determines is appropriate for you, and for whatever reason you're spiking above it too much, and then you get this penalty. Google seems to be more willing to hand out this penalty of late. So if you're in home services and you've got a really successful review plan, I would be careful. **Greg (16:39)** Well, I think Google has been, over the course of the past couple of years at least, focusing more and more on trying to clean up reviews and prevent spam and fraud. And I think one of the reasons for that is because reviews are so important to AI — AI visibility and training — as a source of data. If you use Ask Maps, the content of that review corpus becomes part of the answer for people. Google is more protective of that now and more conscious of the importance of keeping the review corpus as clean as possible, because it's so valuable for AI. That's just a theory. **Mike (17:18)** And they're under more government oversight, particularly in Europe, in regard to reviews. So I think they need to show the enforcers that they're doing something. **Greg (17:27)** Yeah, some compliance, good faith. Okay, so let's shift into Siri search. Apple released iOS 27.2 recently. Siri has been improving gradually over the course of the last number of months, since — I guess June, when they did their developer conference — and it has gotten better. It was lagging ChatGPT and the Google Assistant, of course. It is certainly not at the level of Claude or Google AI or ChatGPT, but it is now much more useful than it used to be, and it can answer a much broader range of questions. So, Mike, why don't you expound? **Mike (18:12)** Sure. As you can see, search is in a default position on the phone. It has been for several versions of iOS; I think we talked about it when it was released. And it now offers much improved local search. It can search not just things like contacts and emails and appointments and tie those all together; it can also search on features and settings and tell you how to run the phone, finally, which is a big help when you're trying to figure something out. But it also does both world search and local search. And searches get saved into the Siri app. **Greg (18:50)** In the Siri app. **Mike (18:51)** In the Siri app. So this was the search I just did just now. **Greg (18:55)** Which is a weird way to manage it, but nonetheless. **Mike (19:02)** And it's a very timely report that came out two days ago, so clearly it's got a fairly fresh index. I think it uses Apple's own web index as a grounding for these world searches— **Greg (19:18)** World knowledge. **Mike (19:19)** World knowledge–type searches. It does a pretty good job on things like local, but also — my wife was talking about blood sugar and potatoes: are sweet potatoes as bad as regular potatoes for glycemic index? **Greg (19:32)** They're not as bad, but almost. **Mike (19:35)** And did you know red potatoes are not as bad as russets for glycemic index? And if you refrigerate them and then eat them, it's even better, right? **Greg (19:44)** Yes, that's right. Resistant starch. **Mike (19:49)** I watch a lot of British police procedurals— **Greg (19:51)** BritBox. **Mike (19:52)** —and there was British slang on there, calling somebody a "wally." What's a wally? I had no idea what a wally was. And I'd never heard of a battered sausage. Woe is me, I've gone to fish places. Apparently all the fish chains basically take a pork sausage, dip it in the same beer batter and deep-fry it. Sounds like a winner to me. Anyway, search is much faster, it isn't bad, and it does a decent job on more complicated searches. So here's a search for "best breakfast near me" — oops, I'm doing this on my keyboard, so it's a little bit wonky, sorry about that. And this has sped up significantly. It's still not as fast as Google, but it's faster than ChatGPT. **Greg (20:54)** So you get a bunch of local listings. **Mike (20:58)** And these are legitimate. These two, for example, are the best. And you can go in then to Apple. Interestingly, there are two tactical issues here. One, you should have your profile in business.apple.com with current photos and current deals, and you should be getting either Apple reviews or Yelp reviews— **Greg (21:22)** Or both. **Mike (21:23)** Or both. Apple reviews are showing up increasingly more. Just thumbs up— **Greg (21:29)** Which are simpler. They're just thumbs up or thumbs down, basically. **Mike (21:32)** Thumbs up and thumbs down, but on a range of things. So you should have photos and Yelp reviews in there as well, and if you can get on the local Guides lists, that's very helpful. The question is whether being world knowledge and local search in your on-phone search — whether that default position will trump years of habit of Google search. It's more private, it's reasonably fast, it's thorough. It doesn't have the deep attributes that Google local search has. It doesn't know inventory, it doesn't know weird food things as well, but it does a pretty good job, and it'll be interesting to see if this cuts into Google's search volume. **Greg (22:18)** As we were discussing in the pre-recording session, there is a pretty large group of people that do use Siri. Siri has been a disappointment for a while, and some of those people have lost interest or faith. I do think there will be a lot of people using this. But as you point out, many of these are informational queries, general knowledge queries, that Google doesn't run a lot of ads against. My belief is there will be some erosion of Google usage, but it won't impact Google in any meaningful way from a revenue standpoint, because of the absence of ads against those kinds of things. What was the score in the game? Is rice or potatoes better for blood sugar? Convert milliliters to the imperial system, all these kinds of general quick-knowledge things. Where can I stream such-and-such a show — I don't know if Siri can answer that last question. **Mike (23:25)** What was the question? **Greg (23:27)** Where can you stream *The Diplomat*, for example. I know that's a Netflix show, but let's see if it can answer that. Yes, Netflix. Okay, great. **Mike (23:40)** From my point of view, the reason I am trying to train myself to use this is that I ask a lot of these very informational questions about technology, about recreation, about food, about recipes. And my experience with Gemini, ChatGPT and Claude is that each of them uses this to profile me in great detail. I don't mind being out on the internet, and with those three I have been asking a lot of questions. But if I can move ninety-five percent of my high-level informational queries to Siri, where it stays on my phone — and anything with the server is totally encrypted, even Apple can't see it — I feel better about giving these guys less data about myself. And I've discussed this previously, but ChatGPT and Claude and Google all did an amazing job of doing a demographic and psychographic profile of me. And it's like, whoa, they know a lot. Who needs it? I guess that's the question I would ask. I don't. **Greg (24:57)** So we'll see. Google Maps was dominant, then a lot of iPhone users switched to Apple Maps and have stayed with Apple Maps, and I'm in that category. I like Ask Maps a lot, and so I use that, but for directions and navigation I always use Apple, and for simple lookups I'll use Apple. If I want to call a business, I typically use Apple. **Mike (25:26)** From a timing point of view, Apple is behind the foundation models and has been criticized for that. But they're clearly playing a different game, and in the jump between 27 and 27.2 — there was no 27.1 for most users — this search is quite good. It depends on the development pattern that they put in place, but clearly they are rolling out improvements to Siri at this point. In two months, it's a major improvement. It'll be interesting to see where they get to. And also, open-source and open-weight models are improving very rapidly. So the question is: is Apple not playing at the frontier going to hurt them in any way? Will, for example, Muse take over the iPhone, or will Apple get there before most users do? **Greg (26:19)** Well, Muse is an agent, which raises a different question. That's a whole different discussion about whether Apple will play in this agentic world— **Mike (26:29)** And the answer is they will. The question is, will they play it soon enough? **Greg (26:36)** To what degree. Google is going to develop a Muse competitor. I mean, it sort of already has— **Mike (26:39)** They have it already. They released it earlier in the week. **Greg (26:44)** I didn't even see that. They released an agent that's comparable to Muse? What's it called? **Mike (26:46)** I don't know if it's comparable, but it's released. **Greg (26:52)** Well, Gemini's had agentic features, but in terms of the representation that "we can do all these things for you — cancel this, make the hotel reservation, do all this planning"… Okay, so maybe they released it earlier in the week; I didn't see that. But regardless, they will have a Muse competitor. And Google is marginally more trusted than Meta is, and Apple is more trusted. You see surveys that say Google is more trusted than Apple. I just don't accept those as accurate. **Mike (27:26)** How could that be? **Greg (27:28)** Well, people don't know enough about Google. But anyway, it'll be interesting to see how this evolves. I think this is a different animal from something like ChatGPT, and I don't think expectations will be quite the same as ChatGPT. People are smart enough to discriminate between these tools, and they're figuring out what to use for what kind of task. And obviously we'll be watching to see how Siri evolves. But let's go on to an agentic feature of Google. Google a while ago introduced something called Ask for Me, which was based on an improved version of Duplex, where Google would call a business and ask for pricing details or other questions. And now they've got one that's directed toward local product inventory, called "Let's confirm it's in stock," or Let Google Call. This is very interesting because, just really briefly: for about 20 years, companies have been trying to bring local, real-time product inventory online, and it's been a very patchy, difficult process, for reasons we don't need to get into right now. Google is kind of the place, through a variety of different surfaces, where you can get access to local or real-time product inventory: local inventory ads, Google Business Profile, Google Shopping. There are a couple of other places where it starts to show up, and this one is arguably the most prominent example. Google has not done a lot to promote those different things, and this is really in your face. You do a local search, or you're looking for where to buy something, and this will show up. So go ahead, Mike. **Mike (29:25)** A couple of things are interesting about these search results. It is a product-driven local search; I'm clearly looking for sneakers in Buffalo. Obviously, as with all local searches these days that are heavily monetized, big sponsored at the top. But it appears here, playing to one of Google's great strengths, which is the fact that they have integrated inventory from a lot of local businesses via Merchant Center— **Greg (29:49)** Right, but also a lot of third parties. There's a third-party ecosystem, like Reserve with Google, that's feeding inventory into Merchant Center or into Google, including Shopify. **Mike (30:02)** Correct. And if a store is feeding inventory to Merchant Center, they get this extra little benny in the organic results. But notice: ads, local results, one organic result, then one Yelp, and then more free listings from their, again, unique inventory database, and then a local listing, but then we're back to sponsored listings. This is a lot of what local results look like these days: a lot of sponsored listings and not many organic listings that you can find. But we're not here to talk about that. We're here to talk about "Let's confirm it's in stock." So if I need a new pair of sneakers, I click this. I'm looking for men's sneakers. I need a size nine wide in a support model; I'm looking for athletic. They didn't ask — well, here we go, now they do. **Greg (30:59)** Right, but they're not asking about your size, right? There it is. There it is. **Mike (31:04)** They didn't the first time I went through, but I'm a nine to ten. Okay: "A list of nearby stores will be sent to you within minutes." They summarize it, they check my email address, you submit it, and then you wait. **Greg (31:25)** And wait and wait and wait. **Mike (31:28)** It said it's been submitted. Let's get out of here and see. **Greg (31:32)** Okay, so interestingly, they do not manage your expectations. You said earlier that they did have something that suggested it would take hours, but in the process that we just saw, they are not managing response-time expectations. They're not telling you these results will get back to you within two hours, ten hours, twenty-four hours, forty-eight hours. **Mike (31:58)** They did finally send something back to me 48 hours after the first time I did it. Over the last week, as I've been testing this feature, it's been literally 48 hours before anything returned. **Greg (32:15)** Well, I did it, and I'm more than thirty-six hours later, and there's just been nothing. Nothing at all. **Mike (32:23)** Right. So in forty-eight hours, we can hang out here and chat till then. **Greg (32:28)** Yeah, people will probably tune out before that expires. Do a telethon, like the old muscular dystrophy telethon. **Mike (32:37)** Yes, we could. As people know, you and I can talk forever. So that's the feature. And send your donations, right. **Greg (32:43)** Yeah. Just look at the 800 number down below and pledge your contribution. Operators are standing by. All right, go ahead. **Mike (32:54)** So unlike Apple, they have this data. This is a huge competitive strength, and it isn't AI-driven, it's database-driven. I don't know, if you go into AI Mode, what you would get. **Greg (33:07)** Well, they're trying to bridge — for small businesses in particular, it's been very difficult to get small-business inventory into this sort of flow. There was a product called Pointy that was very ingenious; it was simple to use, and Google bought that company and integrated its functionality into its larger inventory effort. The big boxes have real-time inventory systems now, where you can buy online and pick up in store, so they have to know what's in store. But this methodology is attempting to bridge the gap between the business listing and what's in store for local businesses. It's a very interesting approach, and now the technology has caught up to the point where you can do this, albeit not particularly effectively yet. **Mike (34:12)** One of the interesting things about the response I finally got with my first test was that it appeared they tried to call 20 businesses. I was looking for engagement rings in a thousand to three thousand dollar price range with a natural diamond. Only one business actually responded to their audio thing. But they had another eight or ten businesses that showed them in Google Merchant Center inventory. **Greg (34:41)** So they're backfilling with that data. **Mike (34:45)** And they make it clear: these were called, these didn't answer, but you might want to call them, because they seem to have it. And then there were ones they didn't know about. So for me, a tactical view here is that even if your inventory is custom or limited, I would create a small inventory via whatever you have — a shop, WordPress or whatever— **Greg (35:11)** Shopify in many cases, yeah. **Mike (35:11)** —and upload it to Merchant Center so that you can be backfilled. Barbara Oliver, for example, refuses to put any of her stock inventory online, and so in any of these inventory-based searches, she gets less visibility. **Greg (35:33)** Well, if you do use Shopify, or one of these third parties that are part of this infrastructure, then they do it on your behalf, so the heavy lifting is done for you. **Mike (35:48)** And you get several other benefits. One, you get this extra little benny here. You may show up under popular products as a local source. So there are other benefits to doing it from an exposure point of view. **Greg (36:03)** Looking forward, this price information is really important to consumers, and availability information is important, and this is something you want to do for AI. You want to have this data out in the world because it's going to be picked up. There is substantial consumer demand to find out where to buy the thing they're looking for in a local store. People know they can default to Amazon; Amazon probably has a version of the product they're looking for. But what they really want is to know where it exists in the stores they're familiar with, or in their area, so they can go and look at it, so they can go get it more quickly. This is a really strategic capability that I don't think Google fully appreciates. **Mike (36:52)** I question, though, whether businesses getting one more call from Google is gonna work. The waters have been so muddied by spammers, and by Google not really legally going after them, that most businesses hear "Google" on the phone and immediately hang up. **Greg (37:12)** Right. I think that's a real problem: they would not trust the automated system calling them. I think they're gonna be pretty resistant to complying with the Google process. "Hi, I'm an AI agent from Google, give me all your information," or "I have a customer who's looking for this item." I think there's gonna be a lack of trust in that, and many people just don't want to be bothered talking to a machine. But it's nonetheless an ingenious way to try to bridge the gap between the data set and demand. **Mike (37:44)** And to try to educate small businesses that there are inquiries about their inventory online. It has this sort of hidden benefit of educating the business. The search is educating not just the consumer but the business. **Greg (37:59)** Well, and Google increasingly calls businesses with AI tools to get information about listings and business hours, to update the listing. So you're right. I think they are trying to condition people to receive these AI calls, to talk to these machines and give over information to them. We'll see how it plays out. It's really clunky right now. The UX that Mike demoed was not too bad, but the long, long wait time is really just not gonna fly. Presumably they will improve it, but it's not a good user experience right now from that standpoint. Okay, so let's move on to our final item today, which is prompted by the Google earnings call yesterday. Google had a rare, as analysts call it, double miss. They missed on the top line and the bottom line. We're not going to do an analysis of all their sources of revenue and what went up and what went down, but they missed on ad revenue by almost a billion and a half dollars. They were under the expectation by 1.4 billion, which is a pretty massive underperformance for Google, given its history of beating most of the time. On the earnings call, Philipp Schindler, who is their chief business officer, said in response to a question, I think from JPMorgan — and this is a verbatim quote — "It's worth noting that the strong growth in AI Mode queries had an impact because of the lower monetization we were seeing in AI Mode." That's really interesting to me, because Google is funneling all these people, with AI Overviews and all its AI Mode promotion on the homepage, into AI Mode. And this is something it wants to do for competitive reasons, obviously. It wants to preempt any defections to Claude or Grok or ChatGPT, and it wants to claw back users that it's lost. **Mike (39:56)** And it wants to train users that this is coming, right? They want to get users used to it. **Greg (40:04)** Right. And so they've been pushing all these people into AI Mode, and it's been working for them. AI Mode is growing, as they've been touting on all their public calls and statements to the press. But what it's done is reduce their ad revenue, which is something people have been talking about for a long time: are they going to be able to monetize AI Mode? And the interesting thing, in my mind, is you've got a lot of SEOs running around saying that in 2027 Google is going to flip the switch and the SERP is going to turn into AI Mode. My response to that is no, and I've consistently said this: no, because they can't get the same kind of revenue out of AI Mode. They are working on it. They've created all these new ad units, some of which are very interesting. But right now they are not monetizing AI Mode anywhere near as aggressively as the regular SERP. **Mike (40:56)** As demonstrated in my sneaker search: ads at the top, ads in the middle. **Greg (41:02)** And for the foreseeable future, they are not going to be able to replace that revenue as they move people over to AI Mode, or some version of AI Mode, some more AI-centric search results. They're trying to manage that whole transition, and it's a tricky thing for them, because their stock, which was down 10% today, at least earlier this morning, is going to take a hit every time they fail to deliver ad revenue. And the Wall Street analysts who recommend the stock are going to lose confidence in Google's ability to manage the transition to AI Mode and to monetize it at the same rate they have the SERP. It'll be fascinating. **Mike (41:43)** Although others have worse problems than Google, right? Google has a lot of balls in the air. They're releasing new TPU hardware; Cloud is growing— **Greg (41:51)** Right. And their Cloud business is growing very fast. **Mike (41:55)** Their search is doing pretty well, all things considered. Their Gemini is doing pretty well in terms of uptake. And their cash flow is being diverted to compete with the massive cash that venture capital is pouring into Anthropic and ChatGPT, and soon the market's going to pour into Anthropic, right? So Google is one of the few companies in a position to, one, compete on so many planes, and two, do so mostly out of cash flow. They have floated some bonds, but they generate so much cash that they can largely outspend and outweigh the expenditures of ChatGPT. In other words, they can spend faster than ChatGPT can grow, at some level. **Greg (42:45)** Right. ChatGPT is the vulnerable one because — even though a lot of businesses are using ChatGPT for coding and other things, Claude is primarily a B2B tool right now, and ChatGPT is positioned as a consumer utility. ChatGPT has a lot of usage: they have 1.2 billion users. I did a back-of-the-envelope comparison, and they're getting maybe nine to twelve percent of Google's daily search volume, which is not huge but not insignificant. So they have a foothold in the market. Google is trying to leverage, as we've discussed many times before, all of its monopoly positions in search and in Gmail and with Chrome, and trying to use all of these tools to get people to use its AI tools, which is something the antitrust case was explicitly trying to prevent. But the judge was such a chicken shit, he didn't do anything with the remedies. And so they are having success migrating people to AI Mode, which eMarketer projected would be bigger than ChatGPT next year. So they're having a lot of success with that. But the dilemma, to come back to this, is that they simply cannot monetize AI Mode at the same level that they have with the SERP. The SERP is so full of ads, and ads that are largely indistinguishable in many cases from organic content — LSAs, the way the PPC ads are positioned, ads in the local pack, a second PPC pack — there's just no way for them to do that in AI Mode unless they want to destroy that experience out of the gate. There's this ramp-up period. This is enshittification in real time. We're watching it, right? **Mike (44:38)** You don't think it'll take the twenty years it took for Google search to enshittify? It'll enshittify much quicker. **Greg (44:41)** No, no way. The pressure's on. **Mike (44:46)** But ChatGPT's problem goes beyond just that. Distribution is a huge issue. Apple clearly has distribution and default position. Google has distribution with Android and search, and default position on Apple for search. Those are huge impediments to ChatGPT breaking through. Anthropic took a different tack: they're primarily B2B and have not lost focus on that. ChatGPT, in being the consumer-known product, is still spending but isn't making it, and I see them, and to a lesser extent Facebook, as vulnerable because they don't own the distribution. Facebook owns more of it. **Greg (45:33)** Well, Facebook has a lot of distribution. Interestingly, with ChatGPT — I'm now paying for the two-hundred-dollar-a-month version of Claude, which is a lot. I'm paying for the twenty-dollar-a-month version of Gemini, and the twenty-dollar-a-month version of ChatGPT. And I don't see the ads on ChatGPT. If I want to see ads, I'll go into incognito mode or sign out and look. ChatGPT obviously is going to more and more monetization; I think they crossed a billion dollars in ad revenue run rate. But interestingly, in ChatGPT, I was doing some research, and a prompt showed up with a little banner above the chat box that said, hey, you should advertise your local business. It was targeting a local business: why not advertise on ChatGPT? I thought, wow, I've not seen this before. I don't want to see too much more of that. But anyway, Google has to preserve a relatively ad-free Gemini and AI Mode experience to keep people using it. If they move a lot of ads into those interfaces too quickly, their usage will go down. Some of the ad units they've showcased look really interesting and potentially really useful, but they can't put nine ad placements in AI Mode. On a SERP you've got, let's call it, seven PPC ads, a local ad, eight ads, maybe LSAs in some cases. So you're looking at 10 or more ads on an individual SERP in many different contexts. And if they were to do that in AI Mode, people would say, "F this, this is just garbage." But they're gonna have to do it quickly as they bring more people over to some version of AI Mode as the primary SERP. **Mike (47:30)** As the Chinese saying goes, we live in interesting times. **Greg (47:34)** For sure. We're watching enshittification in real time of these platforms. **Mike (47:38)** And we're watching diesel engines driving data centers driving us off the climate cliff in real time as well. **Greg (47:46)** That's another terrible conversation, which we don't need to have right now. **Mike (48:01)** Right. So we started with \[Epsom — unclear\], we ended with global warming. Where can we go from here? **Greg (48:05)** Yes. All right. Anyway, next week, hopefully we're going to be having Myles Anderson on to talk about SMB marketing challenges and what they found in their very interesting survey. Like, subscribe, all the things you're supposed to do, tell your friends, and we will see you next time. Interested in sponsoring this podcast or our newsletters please reach out to [mblumenthal@nearmedia.co](mailto:mblumenthal@nearmedia.co) [E-mail Mike ](mailto:mblumenthal@nearmedia.co)