Google ships AI opt-out worldwide as ChatGPT ads pass $1B
Google finished rolling out both halves of its AI visibility toolkit to every site worldwide on 31 August: a report showing your impressions in AI Overviews and AI Mode, and a switch to leave them. OpenAI put a revenue number on the paid layer of the same shift, and the model answering in AI Mode changed again. 6 signals for anyone who wants AI assistants to name their business.
This Week's Signals
Google turns on AI reporting and an AI opt-out for every site
Why it matters for your score
You can now see your impressions in AI Overviews and AI Mode, and exclude your site from them, in the same console. Google states the control is not used as a ranking or inclusion signal for the rest of Search, so being included costs you nothing in normal results. Check how the toggle is set on every property you manage.
ChatGPT ads hit a $1B annualised run rate and open self-serve in more markets
Why it matters for your score
Paid placement inside the assistant is now a funded channel with tens of thousands of advertisers, live in over 40 countries and opening self-serve in India, Europe, the Middle East and North Africa. OpenAI says advertising does not influence the answers, which means getting cited organically is still a separate job you have to do on the content side.
Gemini 3.8 Flash is now answering in Google AI Mode
Why it matters for your score
Google Search VP Robby Stein said the model landed less than 3 weeks after 3.7 Flash. The engine reading your pages changes on a cadence you do not control, so what gets cited can move without a single edit to your site.
Anthropic confirms Claude will watermark the text it writes
Why it matters for your score
Future Claude models will emit watermarked text globally, older models get it over the coming months, and a detection API is already in private preview for regulators, researchers and fact-checkers. If your content pipeline runs on AI drafts, the provenance of that copy stops being your private business.
AIR raises $50M to vet the skills and add-ons that agents call
Why it matters for your score
Roughly 27% of the agent add-ons the company screens get filtered out as risky, and its CEO frames the work as continuous re-verification rather than a one-off scan. If you expose an agent-facing endpoint, expect it to be judged repeatedly, not once.
Google Analytics logged zero traffic for 1 September
Why it matters for your score
GA4 accounts showed a flat zero for that date and Google had not updated its status page at the time of reporting. Annotate the gap now, or in 3 months it reads as a collapse in whatever AI referral trend you are tracking.
Emerging Audit Signals
What to do this week
- 1
Open Search Console on every property you manage and check the Search generative AI control. Google says it is not a ranking signal, so unless you deliberately want out, inclusion is the cheaper setting.
- 2
Export the generative AI performance report for the last 28 days and file it. It carries impressions and no clicks, so the trend line only exists if you start recording it today.
- 3
Annotate the 1 September hole in your analytics before a client or a board deck reads it as a real drop.
- 4
Re-run your crawl access and structured data checks at /tools/ai-readiness-scan. When the model answering in AI Mode changes twice in 3 weeks, the only thing you control is whether your pages are readable at all.
Sources
- ↗ Search generative AI control, Search Console Help
- ↗ Generative AI performance report (Search), Search Console Help
- ↗ A milestone in expanding access to AI
- ↗ Google Search Using Gemini 3.8 Flash In AI Mode
- ↗ How Claude's text watermarking works
- ↗ AIR raises $50M to help companies vet the skills and add-ons AI agents use
- ↗ Google Analytics showing zero traffic on September 1st