Google finished rolling its AI performance reports and generative AI controls out to every Search Console account worldwide on August 31. For the first time, every B2B and SaaS team has a native way to measure its AI search presence and a switch to control it. If you have been inferring your AI visibility from third-party tools, you now have a Google-owned baseline instead, but only if you record it before you start changing anything. No confirmed core or spam update landed this week, so the algorithm section stays quiet.
Week in Brief
- Google rolled out its Search Console AI performance report, and generative AI controls rolled out to every account worldwide on August 31, so any site can now see its AI Overviews and AI Mode impressions and choose whether to appear in those features.
- A new AI citation study, analysed just over 7,000 AI citations on B2B evaluation-stage prompts, found brand-owned pages drew 88.3%, with product pages leading at 24.1% and community sources like Reddit near 4%.
- OpenAI expanded ChatGPT Ads globally and began building out an ads manager, plugins, custom audiences, and conversion matching, as its ad business reached a $1 billion annualised run rate.
- Google confirmed a citation bug tied to Gemini 3.8 Flash in AI Mode and said a fix is coming, which matters for any brand tracking AI citations.
Google’s Search Console AI reports go worldwide
Google finished rolling the Search Console AI performance report and the generative AI controls out to all accounts globally on August 31, confirmed in Google’s own updated documentation, which now states the control reached every website worldwide as of that date. The report shows how your pages perform in AI Overviews, AI Mode, and other AI responses in Google Search, broken down by impressions, pages, countries, devices, and dates, but with no click data. A separate toggle lets you block your content from those AI features. Google says that opting out removes your traffic and impressions from Google and that it is not used as a ranking signal in regular web search.
We covered the initial staged rollout of these Generative AI performance reports back in June when only a subset of properties had access. If your account still isn’t showing the report, our breakdown of why the Generative AI performance report might not appear in Search Console walks through the most common causes.
What this AI report means for B2B, ITES, and SaaS
This is the first time your AI search footprint sits in a Google-owned report rather than a third-party estimate, and that changes what you measure from this week forward. If you run a GEO programme, the impressions view is now your baseline for AI Overviews and AI Mode, the number against which every future content change is judged.
The opt-out toggle is the harder call, and it is not one to make casually. Publishers and tool-led SaaS sites tempted to block AI features to protect clicks should weigh that Google reports no core-search ranking penalty for opting out, while opting out also ends every AI-feature impression and any citation exposure that comes with it. For most B2B sites building for AI visibility, staying in and measuring is the right call. Blocking only earns its place where AI answers are cannibalising a specific high-value page and returning no citation credit in exchange.
Open the new AI performance report in Search Console and record your current AI Overviews and AI Mode impressions as your baseline before you touch anything else. If your Search Console property has yet to get an AI report. Please follow our explainer about visibility of the Generative AI Performance Report.
New research: AI cites product pages most, not Reddit
New research published on September 3 puts a number on where evaluation-stage effort actually pays off. Ten Speed analysed 7,387 AI citations across 170 evaluation-stage B2B prompts, tracked across ChatGPT, Perplexity, Claude, and Gemini. Brand-controllable pages drew 88.3% of citations at this funnel stage. Product pages led at 24.1%, followed by articles at 17.4%, comparison pages at 13.3%, and listicles at 13.2%. Community sources such as Reddit and YouTube together made up about 4.2%, and third-party profiles like G2 and Capterra about 7.2%. Comparison prompts were roughly 20% of the set but produced 26.7% of citations, about 1.33 times their share.
What this research means for B2B, ITES, and SaaS
The surfaces AI reaches for when a buyer is deciding are the ones you already own: product pages and “X vs Y” comparison pages. That reframes the “be everywhere on Reddit” advice, which looks thin at the bottom of the funnel, where community sources drew about 4% in this dataset. A team pouring evaluation-stage budget into Reddit while its product pages still read like brochures is optimising the wrong surface.
The teams most at risk here are the ones with strong top-of-funnel content and neglected product pages, exactly the sites that assumed AI visibility was an off-site game. One caveat keeps this honest: this is one agency’s client base at a single point in time, not a Google-confirmed benchmark. Treat it as directional and test it against your own data; our comparison of AEO tracking tools like Ahrefs, Semrush, and Profound can help you set that up.
OpenAI expands ChatGPT Ads worldwide
OpenAI expanded ChatGPT Ads globally on September 4 and started rolling out ad-stack pieces, including an ads manager, a plugin for integration, custom audience improvements, conversion matching, new ad formats, and budget pacing. Reporting the same week also noted that ChatGPT shopping results appear to draw only from merchants who submit product feeds. Separately, OpenAI’s ad business reached a $1 billion annualised revenue run rate as it opened self-serve buying across India, Europe, the Middle East, and North Africa.
What this worldwide rollout of ChatGPT Ads means for B2B, ITES, and SaaS
ChatGPT is becoming a paid channel, not only an organic citation surface, and the question is no longer whether you will test it but when. If your buyers already research inside ChatGPT, an ad there reaches the evaluation moment that Google and LinkedIn simply miss. The catch is measurement: with conversion matching and audiences still new, early spend is a learning exercise, not a reliable performance line. If your sales cycle is long and multi-touch, treat this as a controlled test, not a channel shift, and hold your budget expectations accordingly.
If your buyers research in ChatGPT, set a small capped test budget for ChatGPT Ads and judge it on assisted pipeline, not last-click leads.
Google clarifies AI Max keyword prioritisation
Google added a paragraph to its AI Max help documentation on September 1, flagged by Search Engine Roundtable, stating that when you disable the broad match campaign setting or move to AI Max, broad keywords are no longer treated as exact match in prioritisation. The document then explains how to manage prioritisation inside AI Max. Google Ads was also seen testing search ads with restrictive match types inside AI Mode on September 4.
What this Update in Keyword Prioritisation means for B2B, ITES, and SaaS
If you have moved campaigns to AI Max, or turned off broad match, the keywords that win prioritisation can shift under this clarified behaviour. A high-intent term that was quietly leaning on exact-match prioritisation may lose standing inside an AI Max campaign. Advertisers with tightly controlled B2B keyword lists are the most exposed here; broad, awareness-led accounts far less so. This is a settings check, not an emergency.
The sealed ad tech remedies
Judge Brinkema’s behavioural remedies stay under seal for about 14 days from September 2. Watch for the specific rules on auction self-preferencing, which could eventually touch Display and Performance Max pricing.
Where AI sends its referral traffic
Similarweb data shared on September 1 showed Reddit and Wikipedia dominating generative-AI referral traffic in July 2026, with Reddit ahead of Wikipedia. Watch whether your category has an active Reddit presence worth earning mentions in and see how this stacks up against where AI Overviews and AI chatbots send clicks more broadly.
Questions We Were Asked This Week
The new Search Console AI report has no click data. How are we supposed to use it?
Use impressions as your baseline, not your outcome. The report shows how often your pages appear in AI Overviews and AI Mode, broken down by page, country, device, and date. Record today’s numbers as a starting line, then track them against each content change you make for AI visibility. For clicks and conversions, pair the report with a rank tracker and GA4, since the AI report will not give you those.
Should we turn on the new toggle to block our content from Google’s AI features?
For most B2B and SaaS sites building AI visibility, no. Opting out ends all your AI Overviews and AI Mode impressions and any citation exposure that comes with them. Google says the opt-out is not a ranking signal for regular web search, so blocking will not hurt your core rankings, but it does remove you from the AI answers your buyers increasingly read.
A new study says AI cites product pages most. Should we stop investing in Reddit for B2B?
At the evaluation stage, shift weight toward the pages you own. Ten Speed’s analysis of just over 7,000 evaluation-stage citations put brand-owned pages at 88.3%, product pages at 24.1%, and community sources like Reddit near 4%. That argues for product pages and “X vs Y” comparison content before Reddit at the deciding moment. Reddit and YouTube may still matter earlier, when buyers are asking awareness questions.
Our AI Mode citations dropped this week. Did our content lose ground?
Check the Gemini bug before you conclude anything. Google acknowledged a citation bug tied to Gemini 3.8 Flash in AI Mode and said a fix is coming, so a fall in tracked AI Mode citations this week may be the bug rather than your content. Confirm the drop against ChatGPT and Perplexity for the same prompts before you act. If citations held on other platforms and only AI Mode fell, you are likely looking at the bug.
White Bunnie Observation
With the AI Performance Report now live across all our client accounts, one of the first things we looked at this week was the page-type breakdown of AI Mode impressions across the B2B, SaaS, and ITES accounts we manage.
An early pattern is beginning to emerge; product, service, and comparison pages are capturing a slightly bigger pie of AI Mode impressions compared with how little of the overall site they represent. At the same time, several blog posts that perform exceptionally well in traditional search are barely surfacing in AI Mode.
So far, the distribution between top-of-funnel and bottom-of-funnel content appears relatively balanced in AI visibility, but the clicks tell a different story. A much larger share of actual traffic is coming through product and service pages.
We are treating this as an observation, not a conclusion. The more meaningful signal appears when a client’s AI impressions are concentrated on pages that attract visibility but convert poorly. That mismatch is now measurable, and it gives us a much clearer direction for where the next round of content and conversion work should focus.
