Brands cited in AI Overviews earn 35% more clicks: what that actually pays, and how to become the cited source

Brands cited in Google's AI Overviews earn 35% more organic clicks than uncited brands — but the figures underneath that headline are a 0.70% click-through rate versus 0.52% (Seer Interactive, September 2025). The 35% is a gain of 0.18 percentage points, roughly 18 extra clicks per 10,000 AI Overview impressions, and knowing that changes which fixes are worth your afternoon.

Sizing the prize before writing the action list is the step almost every version of this stat skips. The reason Contexta works from a site's own Search Console data instead of a generic audit is exactly this arithmetic: 0.18 points is real money on a page pulling 40,000 monthly impressions and a rounding error on a page pulling 400, and only the site's actual query data says which page you're looking at.

What does "35% more clicks" actually pay you?

About one extra click for every 556 AI Overview impressions. Seer Interactive's September 2025 update measured cited brands at 0.70% organic CTR against 0.52% for uncited brands across 3,119 search terms and 25.1 million organic impressions, gathered from 42 organizations between June 2024 and September 2025.

0.70% vs 0.52%

organic CTR, cited vs uncited — this is the "35% more clicks"

7.89% vs 4.14%

paid CTR, cited vs uncited — this is the 91% figure

+0.18 pts

the real organic gap, in percentage points

+3.75 pts

the paid gap — about 21x larger in absolute terms

Two things fall out of those numbers that the headline hides. First, citation does not restore what the AI Overview took: the same study puts organic CTR at 1.76% on results pages without an AI Overview and 0.61% on pages with one, so a cited brand at 0.70% is still running at roughly 40% of its pre-Overview click rate. Being cited is damage control, not recovery.

Second, the paid gap is the big one. A 3.75-point lift on paid CTR is about twenty times the absolute size of the organic lift, which means the strongest commercial argument for getting cited is what it does to the ads you are already paying for on the same query — not the handful of organic clicks it adds.

Is the citation what causes the extra clicks?

Probably only in part, and Seer said so themselves: their researchers noted they cannot prove citation causes the higher CTR, because brands with stronger authority and better baseline click rates are also the brands Google is more likely to cite. The honest reading is that the 35% is an association measured on real impressions, not a lever with a known payout.

That matters for how you spend. If a meaningful share of the gap is selection bias — well-known brands getting both the citation and the click — then a small store that engineers its way into a citation will land somewhere below 0.70%, not at it. Plan for the low end of the range and you will not be disappointed by the invoice.

Which of your pages are worth making citable?

Only the ones with enough AI Overview impressions for 0.18 points to add up — at that rate you need about 5,600 impressions to buy ten extra clicks a month. Most stores have three to ten pages that clear the bar and several hundred that never will, and the whole game is telling them apart before you start rewriting.

Search Console will not label the impressions as AI Overview impressions, so the practical proxy is a page with high impressions, a click-through rate well below what its position should earn, and a query set full of questions and comparisons. That pattern — impressions holding, clicks falling, position unchanged — is the Search Console signature of AI Overview exposure and it is where citable-content work pays for itself.

This is the one part of the job we automated rather than documented: Contexta's Problem Map imports the site's Search Console data and ranks pages by lost clicks per month, so the rewrite queue is ordered by impressions actually at stake instead of by which post you happened to open. It will not tell you an Overview appeared — nobody's tool can, as of mid-2026 — but it does stop you spending a Saturday on a page with 300 impressions.

What actually makes a passage citable?

A section that answers one question completely in its first two sentences, with the number, the date and the source inside the sentence rather than in the paragraph below it. Everything else on the usual nine-strategy lists is real but slower, and ordering the work by cost rather than by prestige is what turns the stat into an afternoon instead of a quarter.

1. 1

   Rewrite the first two sentences under each heading — 20 min/page

   Turns a section you already have into a liftable answer. Best return per hour by a distance.

2. 2

   Rephrase headings as the question a person would type — 10 min/page

   Gives the retriever a sub-query the passage can be matched against.

3. 3

   Move the number, date and source into the sentence — 15 min/page

   A passage carrying a checkable fact beats a passage carrying a claim.

4. 4

   Make the page readable without JavaScript — one-off, a few hours

   No amount of structure earns a citation on a page the crawler renders empty.

5. 5

   Publish original data other people quote — weeks, ongoing

   The strongest signal there is, and the one almost nobody sustains.

6. 6

   Chase brand mentions on third-party sites — months, ongoing

   Genuine effect, wrong first move while your own pages are still unfixed.

The first three lines are the entire quick win, and they work because of how retrieval scores a passage: the model is matching a sub-query against a chunk of text, and a chunk that opens with the answer wins against one that opens with context. We have watched the same section move from ignored to quoted with no new research in it at all, purely from putting the answer in the first sentence and letting the background follow.

The failure mode to watch for is over-correction. Strip a section down to a bare answer with nothing behind it and you lose the depth that made it worth citing in the first place; the answer goes first, the evidence still has to be there.

Do you have to rank in the top 10 to get cited?

No — as of 2026 only about 38% of AI Overview citations come from pages ranking in the top 10 for that query, down from roughly 76% in Ahrefs' July 2025 measurement. Citations are awarded to passages scored against sub-queries, and a page parked at position 14 can own the best passage for one of them.

That is the part that makes the arithmetic above workable for a small store. You are not being asked to outrank an established competitor before any of this pays; you are being asked to write a better self-contained paragraph than they did, which is a fight you can win this week. We unpacked the mechanics of that shift in why top-10 rankings stopped guaranteeing citations.

What is not worth doing to chase AI Overview citations?

Schema markup, mostly — there is no public evidence that Google uses FAQPage or HowTo markup to select AI Overview citations, and Google restricted FAQ rich results to well-known government and health sites back in August 2023. Marking up an unclear answer does not make it clearer to a language model reading the text.

Two more that cost more than they return at small scale: buying a citation-tracking dashboard before you have fixed a single page, and rebuilding your whole content library. Track your citations after you have something worth tracking, and rewrite the six pages that clear the impressions threshold rather than the sixty that do not.

The deeper trap is treating this as a separate discipline with its own budget. The work that earns citations is the same work that earns featured snippets and holds rankings — clearer structure, checkable facts, self-contained sections — which is why we argue that GEO is a writing change rather than a new channel. The 35% is a reason to do it properly, not a reason to start over.