01 The local answer
Google is quoting a lawyer’s own roundup
A local AI summary cited one attorney’s self-promoting listicle as the source for every firm on the list.
Google shows AI Overviews on local results intermittently, and when it does, it writes a short description of each business in the local pack. Local search specialist Joy Hawkins posted a case where Google used a single lawyer’s self-promoting listicle as the primary source for an entire list of law firms, and linked to that page instead of to the firms’ own websites. Barry Schwartz reproduced it on August 12 after several attempts across different browsers.
That is an understudy reading your lines. The description a prospective client sees is written by someone selling against you, and the click that follows it lands on their page.
Two limits are worth stating. This is an observation, not a Google announcement, and it does not fire on every local search, which is exactly why it goes unnoticed. Schwartz needed multiple tries to replicate it, so a single clean search on your own laptop proves nothing.
The underlying mechanic is old: the page that assembled a list of firms outranks the firms it lists. What is new is that the assembly gets read aloud as the answer, with your name inside it and your website nowhere.
What to do about it
Search your practice area plus your city on a phone, signed out, several times over a few days. When a local AI summary appears, screenshot it and note which page it cites. Intermittent features need repeat sampling.
If a competitor’s roundup is describing you, read what it says. A wrong practice area or an implied result is now being spoken on your behalf, and the feedback link on the AI Overview is how you report it. Do not pay for placement on one of those roundups: a page about your firm that you sponsored carries your state bar’s rules on comparative claims and testimonials.
Source
Search Engine Roundtable, Barry Schwartz, “Google AI Overview Local Results Pulling From Low-Quality Listicles,” August 12, 2026 ·
seroundtable.com
· Secondary: Joy Hawkins’s original screenshot
02 The research
Community sites out-cite the press four to one
An analysis of roughly 35,000 ChatGPT citations finds the largest outside source class is the one nobody funds.
Kevin Indig published a citation analysis on August 12 built from about 35,000 ChatGPT citation URLs captured in Profound, covering the United States only, ChatGPT only, and the month of December 2025. Prompts were sorted into four buying stages and citations reduced to root domains, then de-duplicated so one answer citing six Reddit threads counts once.
Set the vendors’ own domains aside and the ranking inverts what most marketing budgets assume. Community platforms hold 17.1% of cited domains, more than four times publishers at 4.0%, with review platforms at 8.6%. Wikipedia, Reddit and LinkedIn account for 99% of that community share, and Wikipedia alone runs between 10.1 and 14.0 points of it depending on the stage of the question.
Think of it as the water table under a field rather than a crop you plant. It sits under every stage of the buying question, it moves slowly, and no amount of spending makes it yours.
The caveats are large, and Indig states them himself: the prompts were software-buying questions by design, which is why vendor domains hold 66.7% to 71.8% of citations throughout. One engine, one country, one month, with 91% of citations in his wider tracking appearing in only a single engine. Nothing here was measured on legal questions.
What to do about it
Do not import the software industry’s source mix into your own market. Spend twenty minutes running the questions your clients actually ask, and write down every third-party domain the answers cite. The source set gets rebuilt per topic.
Where reference-grade sources show up, the only durable move is making accurate primary documentation exist under your name. Never seed reviews or plant threads: a fabricated endorsement is a bar problem long before it is a visibility problem.
Source
Search Engine Land, Kevin Indig, “Community signals are AI’s largest third-party source,” August 12, 2026 ·
searchengineland.com
· Secondary: the underlying G2 analysis and its full methodology, February 13
03 The platform fight
Three hundred French papers want to be paid
Publishers asked France’s competition authority to act on AI Overviews, calling the launch a breach of a 2022 deal.
APIG, a federation of nearly 300 French newspapers, filed a complaint with France’s competition authority on August 11. Google launched AI Overviews in France in late July, and the publishers say it did so without their consent, in breach of commitments Google made in a 2022 compensation agreement. Their president, Marc Feuillee, framed it as a request rather than an objection: “The editors are not seeking to halt innovation.”
This is a bill arriving for a meal already eaten. France’s regulator fined Google €250 million in 2024 over parts of that same agreement, and the EU opened an investigation in December 2025 into whether Google used publisher content to train and run AI services without paying for it.
Google’s position is that the summaries let people ask harder questions and come with tools for publishers to control how their material appears. Both accounts can be true, which is why a regulator is hearing it rather than a court settling it.
It matters outside publishing because the case is about whether being summarised is compensable at all. A law firm’s practice-area page is the same raw material with none of the bargaining power, and no newspaper deal will ever cover it.
What to do about it
Nothing today. Track it, and start keeping your own evidence: a dated monthly log, with screenshots, of what the major AI systems say about your firm. That record is the only proof you will have of what was said and when.
Resist the urge to block AI crawlers as a bargaining chip. Three hundred newspapers can negotiate; one firm cannot, and blocking mostly removes you from answers your clients are already reading.
Source
France 24 with AFP, “French newspapers appeal to antitrust regulator over Google’s AI summaries,” August 11, 2026 ·
france24.com
· Secondary: Search Engine Land’s report, August 11
04 The operational change
Search Console finally shows your AI impressions
The generative AI performance report reached far more properties this week. It still will not show clicks or queries.
Google launched its Search generative AI performance reports on June 3, deliberately to a subset of sites. On August 11 Barry Schwartz found the report live on all 25 or so properties he checks, and Google’s John Mueller confirmed on Reddit that it is still not every domain, appearing only where a site has enough impressions.
What it gives you: how often your URLs appeared in generative AI features across Search and Discover, which pages appeared, plus country, device and date breakdowns down to the hour.
What it withholds is the part everyone wants. There is no click data and no query data, and the report does not separate AI Overviews from AI Mode. It is a guest book with no names in it: you learn that people came, not who they were or whether they did anything.
Even so, this is the first time Google has told a business owner, from its own logs, which specific pages its AI features are pulling forward. That list is frequently not the list of pages you rank with, and the gap between the two is the most useful thing on the screen.
What to do about it
Open Search Console, expand the generative AI tab under the main performance report, set the range to three months and sort by impressions. Put that list beside your top ranking pages.
Where the two disagree, the AI-surfaced pages are the ones to keep accurate, dated and sourced, because a machine is reading them to describe you. A missing tab means Google has not recorded enough impressions yet, not that you are absent from AI answers.
Source
Google Search Central Blog, Hillel Maoz and Moshe Samet, “Introducing Search Generative AI performance reports in Search Console,” June 3, 2026 ·
developers.google.com
· Secondary: Search Engine Roundtable on this week’s expansion, August 11
05 The money
A new toggle aims your ads at people nearby
Performance Max store-goal campaigns gained Local customer optimization, and it refuses to run beside a product feed.
Google published the help documentation for Local customer optimization inside Performance Max campaigns for store goals, and Search Engine Land wrote it up on August 12. The setting prioritises people who are near your business and ready to act right now, reaching them across Google Maps, Waze and local formats on Google Search as they navigate or search for a service in the area.
The restriction is the part that bites. It is not compatible with Merchant Center products or with online conversion goals: you have to deselect the product feed or run a separate store-goals campaign, and adding an online goal later means switching the toggle off first. A linked feed quietly disqualifies the campaign even with the toggle on.
For a firm with an office, this is a dispatcher deciding which calls reach your desk. Which makes the compliance point unavoidable: a local ad carries your state bar’s advertising rules exactly as any other ad does, including comparative claims, results, testimonials and required disclaimers, no matter how much of the assembly Google automated.
What to do about it
If you run Google Ads and have a physical office, decide first which offline action is worth optimising toward: a direction request, a phone call, a walk-in. The toggle appears during setup once you pick offline goals, and under Budget and bidding optimization on an existing campaign. Check the Merchant Center link before you touch it.
Take a baseline before you flip it. Record your current cost per call and per direction request for two weeks, then compare. And route every generated ad asset through the same documented compliance review a human draft would get, with a person signing off before it runs.
Source
Google Ads Help, “About Local customer optimization within Performance Max campaign for store goals,” published August 2026 ·
support.google.com
· Secondary: Search Engine Land, Anu Adegbola, August 12
