9 September 2026 · 6 min read
Google now writes a paragraph above your link and answers the question inside it. You cannot edit that paragraph and you cannot buy a place in it. What you can change is the set of pages it is written from, and most of those pages belong to somebody else.
So getting into an AI Overview is a commissioning job before it is a technical one. The practical answer to how we get in is a list of publications, not a list of fixes to your own site.
An AI Overview is a Search feature rather than a chatbot bolted onto one. Google runs the query, pulls a set of pages out of its live index, writes a short answer from them and links the sources beside it. Two things follow from that, and they pull in opposite directions.
The first is that the pages it quotes are pages Google holds and treats as relevant for that query, which is why the box so often carries a review site, a trade title or a consumer magazine rather than the brands being discussed. The second is that the reader has their answer before they reach a link at all. Several numbers are circulating for how far clicks have fallen since the box arrived and they contradict each other, so take the shape rather than the size: the question is being answered above your link, out of pages that are not yours.
The box is not a ranking of brands. It is a summary of sources, and the two are scored differently. The largest published count of what gets cited says so plainly. Foundation Marketing and AirOps, 5 May 2026, examined 57.2 million citations across 5.1 million AI responses covering 50 brands in seven verticals, spanning ChatGPT, Gemini, Perplexity, AI Mode and AI Overviews, and found that 90% of the sources AI cites when answering are ones the brand does not control; only 10.15% of citations go to brand-owned domains. The study does not state the market its sample covered. Read the study.
Then there is the half of it that stings for a market leader. Surfer, 14 July 2026, took three months of its own tracking across about 5 million unique citation source URLs and 20,000 prompts, again spanning AI Mode and AI Overviews as well as the assistants, and found about zero correlation between a domain's authority score and whether AI cites it, at Spearman minus 0.073 to plus 0.026. The market is not stated. Read the analysis.
You have read the same list three times this month:
It is sensible housekeeping and it makes your pages easier for a machine to read. It cannot do the job you are asking of it, because none of it changes which pages Google pulls in when it composes the answer. A tidier page that was never eligible for the question is still not quoted.
The next move most teams make is to publish their own comparison piece, and that runs into the same wall. A page whose job is to argue that one brand is the right choice cannot be the source for a question about which brand is the right choice. The independence is the whole reason the box quotes publications instead.
IAB Australia and Pureprofile, in the Commerce and Discovery Report 2026, put the question to a nationally representative sample of 1,079 Australians aged 18 to 70 who had shopped online in the past 12 months, alongside 877 New Zealanders, and found that 60% of Australian online shoppers use an AI-powered tool or feature at least sometimes when shopping. Read the report. That figure covers Australian shoppers and nobody else, which is the honest way to hold it up in a British or American meeting.
The categories where this bites first are the ones where a buyer asks a question with several defensible answers: travel, insurance, appliances, telco, energy, business software. A question with one right answer does not need a composed paragraph. A question with six does, and Google will write that paragraph whether or not your brand is among the six.
Anyone promising you a position in the box is selling something they do not control. What you can change is the sources. It takes two things, and most brands have only the first.
The first is a claim worth quoting, by which we mean a claim a journalist could check and a competitor could dispute. A model composing a paragraph about a category needs a specific line it can set against a rival's, and "the leading provider of end-to-end solutions" is not one. The second is that the claim has to be carried by publications Google is pulling into that answer, written by their own journalists under their own editorial standard, because that independence is what made the page quotable in the first place.
Branded content is the mechanism, and we have set out how the citation itself works elsewhere. The work of moving how a category talks about something is not new either: Destination Gold Coast used branded content to shift perceptions of a place people thought they already knew.
One article will not move an answer. The same claim carried by several unaffiliated titles inside one window is what a model has to reckon with, and it then has to be kept fresh, because the index keeps moving. Briefing that many publishers at once is an operational problem before it is a creative one. Our partnerships side exists to run it: one brief, many titles, one window.
What matters here is that it is a buy. Avid commissions across a network of more than 600 trusted media brands in Australia, the UK and the United States, so the publications in the box are ones we can book.
You will not know from a ranking, and you will not know from one blended score. The measurable thing is the set of questions your buyers ask, read across the engines before the work starts and again after it, so you can see which answers now name you, which describe you the way you would describe yourself, and which still belong to a competitor. That is the discipline this channel has always needed and rarely gets: decide what success looks like before the campaign starts.
The place to begin is with what the answers say about you today. We read how AI describes your brand across the questions your buyers ask, name the publications deciding those answers in your category, and give you the read whether or not you take it further with us. That is the AI visibility report. Ask for it before you spend another quarter on your own pages. The argument about what to do next gets much shorter once everyone has read the paragraph a buyer is reading.
10 September 2026 · 6 min read
Before any engine can quote your title, the page has to clear conditions that have nothing to do with the writing. The six that decide it, the two halves of a quotable page, and the crawler distinction most robots.txt files have never made.
10 September 2026 · 6 min read
OpenAI's own documentation says a sponsored placement sits below the answer and cannot change it. That makes the ad and the recommendation two different purchases, out of two different budgets.
9 September 2026 · 6 min read
The largest kind of source in an AI answer is often one nobody can buy. Here is the subtraction that shows what is actually left to commission in your category.
See how your brand appears in AI answers, and where trusted publisher content can strengthen its visibility.
