01 The consumer data
AI names now sit in the brand rankings
YouGov’s quarterly tracker puts Claude and OpenAI beside whisky and camping gear. The trust number tells a different story.
YouGov’s BrandIndex tracks more than 2,000 brands through daily consumer surveys. Its latest Gen Z ranking, published July 29, found Claude’s consideration score nearly doubled in the second quarter, from 14.2% to 28.1%, the highest of any brand in the ranking. OpenAI more than doubled too, from 10.1% to 22.1%, and ChatGPT, tracked separately, grew from 16.3% to 25.4%.
Consideration measures the share of people who would consider buying from a brand next time they are in the market. Johnnie Walker led the list, and REI and Dove Baby posted comparable gains on ordinary seasonal campaigns. Two AI companies now sit on the same shelf as whisky and hiking boots, which is the actual news.
Search Engine Journal’s Greg Jarboe, who analyzed the ranking on August 7, adds the caveat that keeps it honest. Google Assistant’s score also climbed during a quarter when Google was steering users away from it, so some of these gains are name recognition catching a lift, not strategy working.
The counterweight comes from YouGov’s own search research: 28% of Americans say they trust AI assistants for answers, roughly 42 points behind traditional search engines. Consideration and trust are different currencies, and your future clients pay in both.
What to do about it
Your prospective clients increasingly treat AI assistants as household brands, so a recommendation inside one arrives with borrowed familiarity. The trust gap is where an expertise-led firm wins.
Publish what survives a skeptical reader’s follow-up check: named attorneys with real credentials, dated matter results where your bar allows them, and sources a person can verify without taking your word for it.
Source
YouGov, “Brands Rising Among Gen Z in the U.S. in 2026,” July 29, 2026 ·
yougov.com
· Secondary: Search Engine Journal, Greg Jarboe, August 7
02 The research & evidence
A file about cats passed every proof
A joke web standard cleared the exact evidence bar used to sell AI-visibility tactics, in a fortnight.
SEO consultant Mark Williams-Cook got tired of watching “the bots crawled it” and “ChatGPT said it helps” pass for evidence, so he invented cats.txt, a mock web standard for declaring your office cats. He then ran it through the same four proofs cited for llms.txt, a text file widely sold as making websites more visible to AI systems.
The joke file cleared all four: AI bots crawled it, Google indexed it, models repeated its invented facts, and ChatGPT endorsed it as a ranking tactic. Google’s AI Overview even reported the career of a Tuxedo cat named Odd, citing a file describing a cat that does not exist.
The supporting evidence is sober. No AI provider documents using llms.txt for search or discovery, and an Ahrefs crawl of 100,000 domains found the file largely ignored by the crawlers it courts. “None of which is evidence of anything, which was rather the point,” Williams-Cook wrote.
His test is the keeper: ask what would look different if the tactic did nothing at all. Crawled is not read, indexed is not influential, and a chatbot’s endorsement is an echo of whatever the internet said last.
What to do about it
Before any AI-visibility line item lands on an invoice you pay, ask the vendor which result would be impossible if the tactic were doing nothing. If every claimed proof also happens to any text file on the open web, you are buying horoscopes dressed as engineering.
Spend the hour on work you can verify instead: sourced pages under named authors, and the structured data your own server logs confirm the AI systems actually fetch.
Source
Search Engine Journal, Mark Williams-Cook, “How Cats.txt Showed LLMs.txt Evidence Is GEO Astrology,” August 7, 2026 ·
searchenginejournal.com
03 The platform shift
Claude’s words now carry an invisible signature
Anthropic began watermarking AI text worldwide, and detection tools are coming for everyone’s content.
Anthropic said it will embed machine-readable watermarks in text generated by new Claude models, enabled worldwide, fulfilling commitments under the European Union’s AI Act transparency code. The marks are imperceptible to readers and stay with the text through copy and paste, sometimes surviving light editing.
Watermarking happens at the model level, so it rides along whether the text came from the chat app, the API or Claude’s coding tools. Supported image files also receive signed provenance metadata, and Anthropic says detection tools for users and third parties are coming, with technical documentation to follow.
The caveats matter as much as the feature. A detected watermark shows Claude processed the text, not who authored the ideas, and a clean scan proves nothing either, since older models, heavy edits and stripped metadata can all hide the signal.
The mark is invisible ink, and a growing number of readers will own the lamp. Once one major model signs its prose and ships detectors, the question stops being whether AI writing is detectable and becomes who is holding the detector.
What to do about it
If your firm uses AI to draft marketing or client-facing copy, assume some of it now carries a verifiable mark, and that clients, journalists or opposing counsel may eventually run the scan. That is one more argument for the documented human review step this brief keeps recommending: review, edit and stand behind every AI-assisted page under a named author.
Decide your disclosure position in policy, on your own schedule, rather than after someone else’s detector makes the decision news.
Source
Anthropic, “How Claude marks AI-generated content,” Claude Help Center, August 2026 ·
support.claude.com
· Secondary: Search Engine Land, Danny Goodwin, August 11
04 The operational change
Google will dim the ads of unverified advertisers
Limited Ad Serving expands to every Google surface, and identity paperwork is the way through.
On August 5, Google posted an update extending its Limited Ad Serving policy to all Google Ads, adding Gmail, the Play Store and Discover to the Search and YouTube surfaces it already covered. Implementation begins gradually in August 2026 and completes by 2028.
The mechanism is a cap, not a ban. Advertisers Google considers unqualified will have impressions limited in scenarios prone to abuse; individual ads are not disapproved. It works like a dimmer switch: nothing is shut off, your visibility just quietly fades.
Qualification rests on factors you can influence. Google lists advertiser verification status, account maturity, policy compliance history and user reports among them, and its published best practices start with completing advertiser verification and keeping your own brand unmistakable in ad copy and on landing pages.
Affected accounts get an in-account notification and an appeals form, and Google says it cannot promise how long reassessment takes. The cheap move is to qualify before the cap ever reaches you.
Before the rollout reaches you
Complete Google Ads advertiser verification now if you have not. It is the one listed qualification factor entirely inside your control, and it survives whatever order the rollout arrives in.
Then check that every ad names your firm plainly and that landing pages match. For law firms the same discipline serves two masters: state bar advertising rules apply to these ads exactly as they do everywhere else, and the identity clarity Google rewards is the clarity your bar already requires.
Source
Google Advertising Policies Help, “Update to Limited Ad Serving Policy (August 2026),” posted August 5, 2026 ·
support.google.com
05 The money
The ChatGPT ad machine adds parts weekly
Conversion bidding, product carousels, and a data-matching default that switches itself on tomorrow.
OpenAI shipped a bundle of ChatGPT Ads upgrades, reported by Search Engine Land’s Anu Adegbola on August 7. Product feed campaigns gained conversion-optimized cost-per-click bidding in beta, dynamic URL parameters arrived for attribution, and a multi-product carousel format is in testing, with Brazil and Mexico launches queued.
The platform is a store that opened while still under construction, new shelving arriving every week. Measurement hooks now include Triple Whale and Hightouch, and new pixel diagnostics explain why conversion events get rejected. Taken together, it is a performance ad channel assembling itself in public.
One line deserves more attention than the rest. Automatic Advanced Matching, which matches conversions to ads using hashed customer information from your website forms, switches on for existing pixels on August 17, tomorrow, unless advertisers opt out first. This brief flagged the date on August 9; it has now arrived.
For most retailers that is a measurement upgrade. For a firm whose web forms carry the names of people seeking legal help, it is a data-handling decision being made by default, and it deserves a deliberate yes or no rather than a silent one.
Before tomorrow
If your firm runs ChatGPT ads, open Ads Manager today and decide on Automatic Advanced Matching before the August 17 default lands. Hashing does not settle the question: form fields from prospective clients are confidential intake data, and routing them into an ad platform belongs in your privacy documentation, not in a default.
The standing rule carries over: state bar advertising rules apply to ads inside an AI assistant exactly as they apply everywhere else.
Source
Search Engine Land, Anu Adegbola, “ChatGPT Ads rolls out oCPC campaigns, AAM and product carousels,” August 7, 2026 ·
searchengineland.com
· Secondary: Search Engine Roundtable’s summary
