Branded keywords were the safe bucket. AI Overviews now describe your brand for you

Branded keywords were the safe bucket, until AI Overviews arrived#

A fox guards its den by knowing which paths lead to the door, and for years the surest path to a brand's door was the brand's own name. Branded keywords were the quiet, cheap, high-intent bucket: someone types the name, someone clicks the homepage, and nobody writes a strategy deck about it. Then, on 1 October 2026, Search Engine Land published new data from DemandSphere showing Google AI Overviews on more than 80% of the branded queries it tracks by late September, up from about 26% at the start of the month.

The detail matters, because the shape is stranger than the headline. Search Engine Land reported that AI Overviews sat on roughly 27% to 35% of those tracked queries through much of September, then leapt to 69.21% on 26 September and 90.48% on 27 September. The figure stayed high at 80.23% on 28 September and 82.06% on 29 September. DemandSphere tracks branded keywords once a day through its DemandMetrics platform across all markets and devices, and it counted an AI Overview whenever one appeared, whether or not the overview cited the brand.

Two cautions belong on the table before anyone reworks their branded keywords in a panic. First, Google has not announced a change, as Search Engine Land noted, so this is an observed pattern, not a confirmed launch. Second, the chart above is a measure of presence, not of damage. A branded query with an AI Overview on it is not the same as a branded query that lost a click, and the reports say nobody has yet measured that gap. A sensible fox sniffs the wind before it bolts from the hedgerow.

This piece is about the part you can do something about. If you are looking after branded keywords for a company, the job this week is to learn what Google now says about your name, which sources it trusts for the telling, and whether the facts are right. Our work on SEO and GEO starts in the same place: read the page the customer actually sees, for every one of your branded keywords.

Branded queries: the two datasets, and the click gap nobody measured#

Begin with what the evidence is, because there are two datasets and they are not the same kind of thing. DemandSphere's is a daily tracker of branded keywords, so it shows a trend over time. The other comes from Chris Long, co-founder of Nectiv, who tested 100 enterprise brands using SerpApi and, per Search Engine Land, found AI Overviews for 93 of them, including Reddit, Salesforce, Mailchimp and Facebook, and for smaller companies as well. That is a snapshot of one hundred names on one day, not a trend, and the two numbers (80-odd percent of tracked queries, 93 of 100 brands) measure different denominators, so do not stack them into one grand statistic.

Search Engine Roundtable added colour from the practitioner's chair. Barry Schwartz wrote that Google appeared to be showing AI Overviews for almost every large brand name except Google's own name and news publications, quoted Long's line that 93% of the brands he tested had one, and cautioned that it is only about 100 brands. He also observed that most examples had the overview lower on the page rather than at the top, with Adobe a counter-example that did show at the top. His own firm's name did not trigger an overview, although Search Engine Roundtable did, which is a small reminder that this is uneven ground, not a flood.

Read the table slowly and a fox's instinct kicks in: the loudest number is the one with the least to say about money. Search Engine Land put the point plainly, noting that Long warned the expansion could reduce click-through rates from branded searches, but that neither his test nor DemandSphere's data measured clicks. A warning is a hypothesis with good manners. It deserves a place in the plan, and it does not deserve a revenue forecast.

That distinction is why we would not tell a finance director to cut a branded search traffic line by a third this week. If the headline CTR on your branded queries has moved, your own Search Console data will say so soon enough, and Google's AI features documentation says that AI features are included in overall search traffic within the Web search type of the Performance report.

That is useful and incomplete: the documentation does not say how AI Overview clicks are separated from the rest, and the Performance report defines CTR plainly as clicks divided by impressions, per the Search Console Performance report help. Watch the ratio for your own name, week on week, before you trust any borrowed percentage.

There is a human reason to take the signal seriously anyway. A person searching for a name is often at the end of a decision, not the start, and any new box on that page is a new place for the decision to wobble. Whether it wobbles is exactly what nobody has measured, and exactly what your analytics can start to show. Our earlier look at the organic click through rate after AI Overviews took the whole page covers how to set up that comparison without fooling yourself.

Brand name searches and who writes the answer#

The most useful number in the pile is not about presence at all. Slate, a vendor of AI tracking and publishing tools, ran 144 searches for 100 brand names between 30 September and 1 October 2026 and captured every answer and every cited source. It found an AI Overview on brand name searches for 93 of the 100 brands after up to three runs (82 on the first run), and it reported that 71% of the 652 sources Google cited were pages the brand does not own. Slate is selling something here, and its own limitation note says the work is one logged-out scraper in one US location over two days. Treat it as a flashlight, not a survey.

Even so, the shape rings true to anyone who has watched a brand's reputation migrate. According to Slate, Wikipedia and YouTube each appeared in 53% of the overviews, news and financial media made up 15% of all citations, and 13 brands supplied no citations from their own sites. The brand tended to take the first citation and little after it: its own site was the first citation in 52% of answers and the second in 18%. Repeat the identical search and the answer can change, too, since 16 of 44 repeat queries (36%) flipped between showing and not showing an overview.

Here is the sentence to carry into the meeting: for brand name searches, the story about you may be told largely by other people, and in a format that is unstable from one run to the next. A voice on Reddit put the worry in a sentence when it flagged the pattern on 30 September.

That Reddit poster, who also reported noticing the pattern on LinkedIn and said no study existed yet, is a single voice, and the sample was a handful of manual searches. We include it as the mood of the room, not as a measurement. Mood still matters, because it tells you what your own colleagues will ask on Monday about branded keywords. The honest answer to "are we losing control of the narrative" is that for some names, on some days, a large share of the narrative may already be written elsewhere, and the evidence for how many clicks that costs does not exist.

Our own report on the audit that counted 98,020 claims in AI Overviews is the companion read, because a branded overview is only as good as the sources behind it. When a lawsuit, a dictionary entry or a forum post sneaks into the telling of your name, accuracy becomes a brand problem and not just a search problem.

Branded search traffic: what older research can and cannot tell you#

What can the older research tell us about clicks? Some, with care. A widely quoted study of what happens beside an AI summary comes from Pew Research Center, which analysed browsing data from 900 US adults and 68,879 unique Google searches from March 2025. Its headline: users who met an AI summary clicked a traditional result link in 8% of visits, against 15% for those who did not. Pew also reported that users ended their browsing session entirely on 26% of pages with an AI summary, compared with 16% of pages with only traditional results, and that they clicked a link inside the summary itself in just 1% of visits.

Here is the catch, and it is a large one. Pew measured all searches in a single month of 2025, when around one in five Google searches (18%) produced an AI summary, and it did not isolate branded queries. A person searching a name has a different job in mind from a person asking a question, so the gap for branded search traffic could be smaller, equal or larger. The research gives a reason to monitor your branded keywords, not a number to plug in. That is the discipline we would apply to any borrowed benchmark.

Google, for its part, takes the opposite view of the same pages. Google's AI features documentation says that when people click from results pages with AI Overviews, those clicks are higher quality, meaning users are more likely to spend time on the site. That is a company statement about quality, not a published rate of clicks, and it does not contradict Pew: fewer clicks can be better clicks. A brand with a trial sign-up or a basket needs to know which of the two it is getting, and only conversion data from its own analytics can say.

There is also a cost-benefit instinct worth keeping in check. Google's robots meta tag documentation says that nosnippet will also prevent content from being used as a direct input for AI Overviews and AI Mode, and max-snippet limits how much can be used. Those controls exist, and the temptation to reach for them is understandable. They also shrink your ordinary snippet, so an opt-out may trade a risk you cannot yet size for a loss you can. We covered that trade-off in our look at the opt-out button that quietly removes you from the news.

Monday morning: audit, fix the facts and keep ai mode seo on the horizon#

So what do you do on Monday? Start by getting the facts on one page. List your branded keywords: the exact name, the name plus pricing, plus login, plus reviews, plus alternatives, plus the founder's name, plus the product names people actually say. Search each from a clean browser profile, three times on different days, because the Slate repeat test showed answers flipping. Record whether an overview appears, where it sits, what it says, and which sources it cites. That sheet is your audit of branded keywords, and it costs a morning, not a budget.

Fixing the facts means working in the places Google says it listens. Its Organization structured data guidance says that adding organization structured data to your home page can help Google understand your organisation's administrative details and disambiguate it in search results, and can influence which logo appears in Search and the knowledge panel. Where a knowledge panel exists, the knowledge panel help page says verified users can suggest an edit through the Feedback option on search results pages. None of this guarantees what an AI Overview will say, and Google's documentation does not promise it. It does tidy the trail the machine follows, and an untidy trail is how a fox gets mistaken for a dog.

The second job is to look at who else is writing about you. If the Slate pattern holds for your name, Wikipedia, YouTube and news pages are doing a good deal of the talking, so the facts in those places matter as much as the facts on your homepage. That is brand work as much as search work, which is why our brand strategy team and our content marketing team end up in the same room as the SEO team. For a sense of how a family of names behaves when it shares a search results page, see our companion piece on brand portfolio strategy for BT, EE and Plusnet.

The third job is paid. Branded keywords are often where an advertiser's cheapest conversions live, and an overview on the page may change how people scroll. If you run branded campaigns, compare your branded impression share and click-through rate this fortnight against the last, and read our piece on why a Google Ads landing page destination stops working so you can rule out ordinary breakage before you blame the machine.

Finally, keep one eye on the next room. Search Engine Journal reported that Google had expanded information monitoring in AI Mode, with Robby Stein announcing on X that you can tell AI Mode exactly what to look for and Search will keep checking sites, forums, social posts, real-time data and the Shopping Graph. We mention it as a related change, not as this article's subject: if people can ask a machine to watch your brand across the web, then the pages and profiles it reads will matter even more. Treat ai mode seo as a neighbouring question to answer after the branded audit, not before it.

If you want a second pair of eyes on the audit, our AI consultancy desk builds exactly this ledger, and our earlier look at zero click searches and the pennies AI Overviews pay shows the same discipline applied to publishers. The pattern repeats: measure first, then move the budget.

Frequently asked questions#