When a brand search shows an answer you did not write

Search your brand name before you read another chart about it.

Do it logged out, on desktop and on a phone. Then search a common misspelling, and the names of the one or two products people actually type. Save the generated text and every URL the summary cites. That folder is the piece of work. The September headlines are context for it.

Two questions, not one

A brand query and a category query are different jobs.

On a category prompt, the question is whether you show up at all. On a search for the brand name, the homepage can still be on the page. Appearing is not the useful test. The questions that matter are whether the summary is right, and whose pages it cites.

A summary can name you and still describe a product you retired, a market you left, or a positioning line nobody on the current site would sign. That is a different failure from being absent.

What changed, without hanging on one number

In the last week of September 2026, AI Overviews started showing up on a much larger share of brand-name searches. Google did not announce it.

Three samples pointed the same direction. DemandSphere's daily series of branded keywords jumped into the 80s by the end of the month, after sitting near a quarter of keywords on September 1. Ahrefs, on a large US desktop crawl, ended September 29 at a similar level, up from the low 60s in July. A hand check of large brands, reported by Search Engine Roundtable, found an AI Overview on most of the names checked. News publications and Google's own name were among the gaps.

The exact percent is the wrong thing to publish as a target. DemandSphere's series moved by tens of points within a couple of days, including a drop after the peak. Direction and timing are the finding. A single screenshot from one afternoon is not a baseline.

Where it sits is not settled

Ahrefs saw most bare brand-name overviews in the top position, above the company's own listing. Search Engine Roundtable, looking at a hand check of large brands, saw many of them mid-page or under the site snippet.

Treat placement as a test. Google changes these layouts often. The claim worth keeping is narrower: a brand search can now show an answer you did not write, and that answer can sit where the homepage used to be the whole story. It does not always.

Who writes the summary

Ahrefs looked at the sources cited in the branded overviews it could inspect. Wikipedia showed up most often, then YouTube, LinkedIn, and Yahoo Finance. Across those source lists, one verified competitor-owned site appeared.

Two readings, and only one of them is comfortable. Competitors are not, in that sample, taking over the summary. The description a searcher may read first is still mostly assembled from encyclopedia entries, video, a professional network, and finance pages. Those pages can be right. They can also be years out of date.

The missing number is the one to collect yourself. The published counts do not say how often the brand's own site is among the cited URLs. "We appear in the overview" and "our pages are what the overview used" are not the same fact.

Fix the sources, not the results page

You cannot edit the overview. You can edit, or request edits to, the pages it is likely to read.

Start with the sources that showed up most often in that Ahrefs list: the Wikipedia article if one exists, the LinkedIn company page, the YouTube channel, and any finance or retailer listing that still describes the company. On Wikipedia, follow the conflict-of-interest guidance and propose the change on the talk page. Do not quietly rewrite your own article.

Then give the summary a first-party sentence it can use without inventing one. An about page that says what the company does, who it serves, where it operates, and what it sells is enough. Keep that page consistent with the third-party profiles. Contradictory sources are how a model picks up a retired product line and states it with confidence.

Repeat the logged-out check on a schedule. The public series moved sharply inside a week, so one capture is a starting point, not a trend.

What Search Console can and cannot tell you

Split branded clicks from the rest of the Performance report. Google added a branded queries filter in November 2025. Mark the last week of September on the chart. That is when the third-party series turned up, not a date Google documented.

If branded clicks fall while impressions stay flat, that is a clue that the results page is holding people longer. It is not proof that an AI Overview caused it. Search Console does not show AI Overview appearances by query. Pair the chart with the screenshots and the cited URLs, or you will explain a measurement change as a behavior change.

What this is not

This is not a ranking-system change, and Google has not said that it is. It is not a new discipline called optimizing for AI Overviews. It is not a reason to buy a visibility score and call the work done.

The same third-party pages already feed other answer surfaces, which is the practical point of generative engine optimization: be citeable, and be accurate where you are cited. The new part is narrower. Those pages can now shape the query you used to treat as yours.