Are AI Citations Worth Anything?

Two respected studies appear to disagree about whether AI citations matter, and the disagreement is where the real answer lives. We run three production builds and route AI-assistant referrals into a dedicated GA4 channel on our insurance build, so this page reconciles the published numbers with the instruments we actually operate.

How often do people actually click AI citations?

Almost never — about 1% of the time an AI summary is on screen. Pew Research's analysis of March 2025 browsing data, published July 2025, found Google users clicked a traditional result on just 8% of visits that displayed an AI summary, versus 15% of visits without one — and clicked a link inside the summary on roughly 1% of summary visits.

The direction is corroborated elsewhere. Ahrefs measured a −34.5% CTR drop for top-ranking pages on queries where AI Overviews appeared (March 2025), and Seer Interactive's first CTR study found organic CTR on Overview queries fell from 1.41% to 0.64% (January 2024 vs January 2025). If your entire model of citation value is "citations send clicks," the evidence says the model is wrong.

Do cited sites get anything back?

Yes — a measurable relative advantage. Seer Interactive's 2026 update, covering 53 brands from January 2025 to February 2026, found that being cited in an AI Overview delivered 120% more organic clicks per impression than not being cited — though still 38% below queries with no Overview. Where the brand was not cited, organic CTR on Overview queries declined 67% over 2025. The same update recorded Overview-query CTR hitting a 1.3% floor in December 2025 and climbing to 2.4% by February 2026, so the decline is not fixed.

How can the 1% and the +120% both be true?

Because they measure different events. Pew counts user clicks on the summary's own links — a behavior almost nobody performs. Seer compares a brand's organic clicks per impression when it is cited in the Overview versus when it is not — and there, being the named source correlates with more than twice the clicks per impression in a shrunken click pool.

Reconciled, the model looks like this: an AI citation is a weak pipe and a useful marker. Few users travel through the link. But the engines tend to select sources that already exhibit the structure and authority searchers reward, and appearing as the named source may confer an additional advantage. Correlation deserves its usual warning — Seer's data cannot prove the citation causes the higher CTR rather than accompanying it.

Do AI referral visitors convert better?

The published answer is a caveated yes; our own answer is "not yet measured, on purpose." Semrush's AI referral research reports AI-referred visitors converting 4.4x better than average, and that number should never travel without its survivorship caveat: sites that measure AI referrals are, by definition, sites already visible in AI answers, so the multiplier describes the winners.

Our first-party instrument is younger than the question deserves. We run a dedicated AI-assistant referral channel behind a first-party tag gateway on our insurance build, shipped in August 2026 [our data]. Its early data already shows the volume story — a thin slice of sessions next to organic — but we will not publish conversion quality from days of data. When the channel has real history, the finding ships here, whichever direction it points. The setup for measuring this yourself is documented in how to track AI traffic in GA4.

What is a citation worth when nobody clicks it?

Presence and accuracy — the value that never appears in a sessions report. When an answer engine cites your page, your framing and your numbers shape the answer thousands of people read without visiting anyone. When it cites someone else, their framing shapes what those people believe about your category, and any errors about your brand go uncorrected.

That is why we treat citation work as brand infrastructure rather than a traffic channel, and why the one citation we can fully document — observed through dated manual checks within days of the page shipping — is written up as a field report, not a victory lap. The mechanics of earning that kind of selection are covered in our GEO guide.

How do you measure what a citation is worth to you?

Run a 90-day measurement before trusting anyone's average — citation value is distributed too unevenly for an industry mean to price your case. The protocol costs a spreadsheet and a recurring calendar slot, not a budget line.

Three instruments cover it. First, a GA4 channel that isolates AI-assistant referrals, so the sessions and conversions those visitors produce stop hiding inside generic referral traffic. Second, a monthly prompt battery: the same category questions, asked the same way, in each engine, recording which sources get cited and whether you are among them — the free method is documented in measuring AI share of voice without tools. Third, branded queries in Search Console, because a reader who saw your name in an answer often returns by searching for you rather than clicking the citation link.

If 90 days shows citations without conversions and no branded lift, that is a real answer too. Reallocate the effort — the measurement cost you almost nothing, which is the point of running it before the budget conversation.

When is chasing citations not worth it?

When your economics depend on the click — and for many sites, they still do. A publisher monetizing pageviews cannot invoice a 1% citation-click rate; an affiliate site loses the commission when the assistant summarizes the comparison. If that is your model, defending and converting the clicks you still get is the better use of the same hours, and is SEO dead covers what we are reallocating and what we are not.

Citation work earns its keep in the opposite cases: brands whose buyers ask assistants for recommendations, categories where wrong AI answers cost sales, and sites whose conversion happens off a handful of high-intent visits rather than volume. Run your own numbers for 90 days before believing anyone's multiplier — including ours.