How to get cited in Google’s AI Overviews

Most marketing teams are still measuring AI Overviews wrong. They check whether their page ranks, see a summary box has eaten the top of the SERP, and assume the fix is more content. It isn’t. Inclusion in an AI Overview is an entity decision, not a page decision, and the brands getting cited are the ones search systems already recognize as credible sources on a topic.

That distinction matters because the click economics have shifted. SparkToro’s 2024 analysis found that only 360 of every 1,000 U.S. Google searches lead to clicks on the open web, and 374 of 1,000 in the EU. More than half of all sessions end inside Google’s interface. If your brand isn’t named in the answer, the session ends without you.

Citation is now a performance metric

Seer Interactive studied 3,119 informational queries across 25.1 million organic impressions and 1.1 million paid impressions. The pattern they found is hard to ignore. When no AI Overview appeared, organic CTR averaged 1.45%. When an AI Overview appeared and the brand was not cited, CTR collapsed to 0.52%. When the brand was cited inside the Overview, CTR recovered to 0.70%, and paid clicks rose 91% compared to the non-cited condition.

Read that again. An uncited Overview cuts your organic CTR by roughly two-thirds. Citation doesn’t just add visibility; it stops the bleed.

Seer is careful to note that correlation isn’t causation. Brands that get cited may already have stronger baseline authority. Fair point. But that’s the whole argument: the brands winning citation are the ones with documented entity strength, and that strength is buildable.

What Google’s own documentation tells you

Google’s Search Quality Rater Guidelines instruct human raters to perform formal “reputation research” on websites and creators. Raters are told to look beyond the site itself, at news articles, Wikipedia entries, blog posts, magazine pieces, forum threads, and external ratings. Google says raters don’t directly change rankings, but their feedback trains and refines the systems that do.

That’s the loop most SEO teams ignore. If raters are evaluating you in publications a rater would examine, your off-site footprint is part of the quality signal, even if it never throws a backlink.

Google also states, in its helpful content guidance, that trust is the most important factor inside the experience, expertise, authoritativeness, and trustworthiness framework. The self-assessment questions Google publishes for site owners ask whether a site is “widely acknowledged as an authority” on its topic. Widely acknowledged. By whom? Not by you. By independent sources.

The entity, not the page

Generative engine optimization (GEO) operates at the entity level. Traditional SEO improves a single URL’s position in organic listings. GEO increases the probability that a search system selects your brand as a source while assembling an answer. Those are different problems.

A page can rank well and still be skipped during summary generation if the brand behind it lacks recognized authority on the topic. The reverse is also true. A modestly ranked page from a recognized entity can be cited because the system trusts the source.

This is where most B2B programs underinvest. They publish content, track keyword positions, and never measure whether the underlying entity is gaining ground. Branded search demand, citation rate, and independent reputation coverage tell you that. Rankings alone won’t.

Building citation-worthy authority

A few things move the needle here, and none of them are quick.

Branded search demand. SparkToro analyzed 331,697,810 Google searches over 21 months and found that just over 44% of queries included a branded term. Branded queries signal prior awareness, which signals recognition, which feeds the entity model. Track branded search volume monthly. If it’s flat, your top-of-funnel work isn’t producing recognition, regardless of what your traffic dashboard shows.

Independent reputation coverage. Get referenced in the kinds of sources Google’s raters review: industry publications, association directories, analyst write-ups, podcast appearances with reputable hosts, Wikipedia where eligible. A single placement in a respected industry publication tends to outperform a dozen low-tier guest posts (the kind of guest post network most agencies still sell, frankly, is a waste of budget).

Consistent authorship and subject depth. Google’s E-E-A-T self-assessment asks whether content has a clear author and whether that author demonstrates relevant experience. Anonymous content, or content under a generic “marketing team” byline, is harder for systems to attach to an entity. Name your authors. Build their topical track record. Link their bios to their work elsewhere.

Topic-level breadth, not just depth. Seer’s data shows that brands cited across several related queries demonstrate steady selection during answer assembly. One pillar page won’t do it. A documented body of work across the topic cluster will.

Structured data and clean technical foundation. This is table stakes. If a search system can’t reliably parse who published the content, who wrote it, and what entity it belongs to, citation becomes harder. We see this constantly on WordPress sites with five years of plugin sediment. Schema is broken, author entities aren’t connected, organization markup contradicts the About page. Fix the foundation before chasing citation.

This is the part of the problem 321 spends most of our time on with clients. Custom WordPress builds with clean entity architecture, author schema wired correctly, content programs that publish with depth across a defined topic set, and attribution setup so you can see citation rate move alongside branded search and pipeline. Citation is the output. The inputs are technical, editorial, and reputational, and they have to be managed together.

Measuring it

Stop reporting on AI Overviews as a binary “are we in or out” question. Build a tracked query set, somewhere between 100 and 500 informational queries that matter to your buyers, and measure citation rate as a percentage. Watch it monthly. Pair it with branded search demand from Google Search Console and a documented log of independent reputation placements.

Edelman’s 2026 Trust Barometer, which surveyed 33,938 people across 28 countries, found that 62% of respondents who trust influencers said they would reconsider a company they currently distrust if a trusted voice endorsed it. Trust transfers. The same mechanism is at work when a search system decides whose content to put inside an answer. External validation is doing the heavy lifting.

A practical next step

If you’re reading this and your team has never measured citation rate, start there. Pull 100 informational queries from your topic cluster, check them weekly for AI Overview presence and citation, and log the results in a sheet. Within a month you’ll know whether you have a citation problem or a recognition problem, and they have different fixes.

If you’d like a second set of eyes on the entity architecture, content depth, and reputation footprint that feed AI Overview citation, that’s the kind of work we do with mid-market B2B teams. Happy to compare notes on what’s working in your category and where the gaps usually sit.