August’s State of Search & AI covers why being cited isn’t the same as being recommended, what a million keywords reveal about search demand, and more.
August’s State of Search & AI covers why being cited isn’t the same as being recommended, what a million keywords reveal about search demand, and more.
New research reveals that appearing as a cited source in an AI answer and being the brand AI recommends are two very different things — and that conflating them leads to a false sense of visibility. At the same time, a large-scale keyword study offers some of the most specific data yet on which categories are actually losing ground to AI and which ones aren’t. The picture is more nuanced than the zero-click narrative tends to suggest.
July was the month AI visibility finally became measurable. Google introduced AI impression data in Search Console, while Microsoft launched an AI Citation Share metric in Bing Webmaster Tools. July also put a concrete number on zero-click search: 68% of U.S. Google searches ended without a click in early 2026, up from 60% in 2024.
August moves from “can we measure AI visibility” to “what does that visibility actually mean” — and the answer is more complicated than most dashboards will show.
This month’s research cuts through some of the more optimistic assumptions about AI search performance. Citation counts don’t equal recommendations. Local map pack rankings don’t translate to AI recommendations. And AI’s impact on search volume is real but uneven depending on brand category.
AI Mode now runs on Gemini 3.5 Flash, the model Google introduced at I/O as the engine behind the most significant search redesign in roughly 25 years. The upgrade matters for content strategy because it carries conversational context across follow-up questions and is more selective about which pages it trusts enough to cite. Thin location pages, generic service descriptions, and content that covers a topic without adding genuine depth are at higher risk of being passed over.
What this means for marketers: Content audits should go beyond keyword mapping and ask whether individual pages offer something substantive. Generic coverage of a topic is a weaker position than it used to be, and that gap is likely to widen.
Independent testing across roughly 165 London businesses and 72,000 questions about UK retailers found that AI tools returned at least one false fact (most commonly, an incorrect postcode or address) for the majority of multi-location brands tested. The errors were basic information a customer would check before visiting a location.
ChatGPT, Gemini, AI Overviews, and Perplexity don’t notify brands when they surface incorrect information. The only way to catch errors before customers do is to manually query each platform with a defined set of location-specific questions and check the answers against ground truth.
What this means for marketers: For multi-location brands, this is a direct operational concern. Incorrect addresses or hours in an AI answer cost a customer visit, and the brand has no way of knowing it happened unless someone is checking. A lightweight recurring audit is worth building into regular local marketing operations.
Research found that when a brand’s own content was cited as a source inside an AI Overview, that brand was still left out of the model’s actual recommendation 69% of the time. In those cases, Google read the page, extracted information from it, and then recommended competitors mentioned within it — not the brand whose page it was.
A lot of emerging AI visibility tools track citation counts as the primary success metric. If citation doesn’t reliably predict recommendation, and recommendation is what drives discovery and consideration, then optimizing for citation alone is optimizing for the wrong thing. The more meaningful metrics are recommendation share (how often your brand is the one named), brand accuracy (does the model have your basic facts right), and presence (how often does your brand appear by name in responses to relevant queries).
What this means for marketers: Audit what your AI visibility reporting is actually measuring. If the primary number is citation count, it’s worth adding recommendation tracking alongside it. That distinction will matter more as AI search matures.
A study of over 50,000 commercial keywords found that AI Mode showed text ads on 29% of commercial searches overall, rising to 54% for keywords with CPCs of $10 or more. That’s a meaningful ad surface inside an AI experience, and it signals that Google is finding ways to monetize AI Mode at scale.
Only 11.5% of advertiser domains were also cited on the same queries where they ran ads, and just 2.3% of advertised URLs ranked organically for those terms. Paid, cited, and organic are three largely non-overlapping channels in AI Mode. Running an ad on a query doesn’t improve your chances of being cited or ranked organically for it.
What this means for marketers: Paid and content strategy need to be planned as separate levers in AI Mode. A paid placement captures a click from users who see the ad, but doesn’t build the kind of credibility that earns a citation or an organic recommendation.
A study compared AI recommendation rates to Google local 3-pack appearance rates for the same set of brands, and the gap is striking. The brands in the study appeared in Google’s local 3-pack 35.9% of the time. ChatGPT recommended those same brands just 1.2% of the time. Gemini recommended 11%, Perplexity 7.4%. This implies that traditional local SEO performance and AI local recommendations operate on almost entirely different signals.
The research also proposes a practical audit framework for understanding local AI visibility before investing in local content. The approach involves running a defined set of prompts — discovery queries, comparison queries, trust queries, and logistics queries — across each major AI platform, then scoring the mention rate, factual accuracy, and the way the brand is framed. Technical eligibility gets fixed first, trust signals come second, and content depth comes last.
What this means for marketers: A strong local 3-pack presence is not a reliable indicator of AI discoverability. Brands that have invested heavily in local SEO should run an AI-specific baseline audit to understand where they actually stand.
A study of over one million keywords found that 29% of search volume is declining and 20% is growing, with the net effect being roughly flat overall. But the distribution matters. The categories losing search volume fastest are the ones AI can answer completely within the chat interface, like health information queries, financial definitions, and how-to explanations. The categories holding steady or growing are the ones that require a transaction, a comparison across specific options, or a decision that AI can inform but not complete.
The pattern makes intuitive sense. If someone asks an AI how to treat a minor injury, they may never need to visit a health website. If someone is comparing rental furniture options or booking travel, the AI answer creates downstream intent that ultimately leads to a website visit or a transaction.
What this means for marketers: The zero-click and AI absorption narrative is real, but it’s not evenly distributed. The more urgent task is understanding which specific query types within a category are declining, which are stable or growing, and calibrating content investment accordingly.
August’s research corrects some of the assumptions that have been built around AI search optimization. Citation is not a recommendation. Local rankings are not AI discoverability. Paid placements don’t transfer to organic or citation lift.
At the same time, the keyword volume study is the most grounded reassurance yet for brands in transactional categories. AI isn’t absorbing the search queries that lead to purchases, rentals, or high-consideration decisions at the same rate it’s absorbing informational ones. The commercial signal in search is more durable than the aggregate zero-click figures imply.
For marketers, that means:
You need to ensure you look at the right metrics when examining discovery visibility. Citation counts, impression totals, and map pack rankings are all telling a partial story. Recommendation share, factual accuracy, and downstream conversion are where the full picture lives.
August’s State of Search & AI covers why being cited isn’t the same as being recommended, what a million keywords reveal about search demand, and more.
E-E-A-T shapes Google rankings and AI citations alike. Discover the expertise signals that help your brand earn trust from search engines and AI platforms.
If rankings tell one story, search share of voice tells the whole plot. Learn how to measure SOV across SEO, PPC, and AI-generated search results today.
