By Priya Raman, AI SEO, GEO & Martech Analytics. Last updated: 2026-11-08
What is the difference between a mention and a citation?
A mention means the engine named your brand in the answer text. The engine can do that from general knowledge, from training data, or from pages it read, and no page of yours needs to be involved. A citation means one of your pages was listed as a source for the answer. That requires the engine to have retrieved, or at least selected, your page for that specific response.
The two events reward different work. Mentions reward brand presence: being talked about, reviewed, listed, and named across the web in contexts the engine absorbs. Citations reward page-level usefulness: a specific page that answers the prompt's question well enough to be used as evidence. A team can be winning one race and losing the other, and the work that fixes each is different. That is why the distinction matters before any numbers enter the discussion.
What did the MarketScale study actually measure?
The figures in this post come from one source, labelled wherever they appear: the MarketScale State of GEO, AI Visibility for B2B Companies study, read on 2026-11-03. Its reporting window ran from August 4 to November 2, 2026, with the snapshot updated November 2, 2026. It covered 169 projects across five AI engines, running 11,934 prompts and collecting 552,661 responses. It is a vendor study, published by a company that sells in this category, so treat its figures the way you would any vendor benchmark: as a large, dated sample that shows the shape of the phenomenon, not as settled fact about your market.
Across the five engines, the study's unweighted averages were a 24.84% mention rate and a 17.98% citation rate. The averages already tell the first half of the story: even before engine differences enter, mentions and citations are not the same quantity, and a score averaging them is averaging two unlike things. The engine-level figures tell the second half. Google AI Mode named brands in 20.9% of its responses but cited brand pages in only 5.8% of them, the widest gap of the five engines in the study. An engine can clearly know a brand, talk about it freely, and still rarely use that brand's own pages as sources.
Want to put this into practice?
Reachium automates LinkedIn outreach, content publishing, and inbox management in one platform.
See Reachium →Why does a blended score mislead a GTM team?
Because it converts a diagnosis problem into a morale number. Suppose your blended score rises. Three different realities could sit underneath: mentions rose while citations fell, citations rose while mentions fell, or both moved together. The aggregate rates alone do not say why either rate moved; that takes reading the actual answers. Each reality calls for different work, different budget, and a different conversation with leadership. The blended score supports none of those conversations. It supports one: whether the number went up or down, which is the least useful question in the room.
The blended score also hides engine strategy. If one engine drives most of your AI-referred pipeline, its mention and citation rates are the ones that matter, at their own values, not diluted by four engines your buyers rarely open. Per-engine reporting is how a team finds out which engine is actually sending it business, and our own tracking work on this portfolio starts from exactly that question: who comes from what, attributed by engine, before any portfolio-level claim gets made.
How should a GTM team report AI visibility instead?
Four habits, all cheap. First, fix a prompt set that mirrors your buyers' real questions and version it when it changes, so periods stay comparable. Second, log mentions and citations as separate columns, per engine, per run; the sampling method in how to measure AI search visibility covers how to run this without fooling yourself. Third, report the pair, never the blend: for each engine, mention rate and citation rate side by side, with the gap itself treated as a finding. A widening gap on an engine that matters is a prompt to inspect the actual answers before diagnosing a cause: the rates show the gap, not what produced it. Fourth, cross-check against referral analytics. Citations that never produce visits may still build familiarity, but a citation series with no referral floor underneath it deserves a sceptical read before it drives spend.
None of this requires waiting for a perfect tool. It requires refusing the single number when a vendor offers it, including when the vendor is your own dashboard.
FAQ
Is a high mention rate with a low citation rate bad? Not automatically, but it is unfinished. It means engines know your brand exists and describe it, while sourcing their answers elsewhere. The pages being cited instead of yours are the most useful competitive research available to you: they show the structure and evidence the engine preferred on your buyers' questions.
Should I ignore blended AI visibility scores entirely? Use them the way you would use a weather icon: a rough glance, never a plan. The moment a score informs budget, content priorities, or a board update, split it into per-engine mention and citation rates first. If a tool cannot show you those components, that limitation belongs in every report the score appears in.
Which engine should a B2B team prioritise? The one your buyers use and your analytics can see. The MarketScale figures show engines behave differently, but they do not tell you where your pipeline comes from. Your own referral data, read per engine, answers that. Portfolio teams should expect the answer to differ by site and audience, which is another reason a single cross-engine score cannot carry the decision.
How often should mention and citation rates be rechecked? On a fixed schedule you can sustain, with repeat runs inside each period, because single runs are noisy. Trend over several periods is the signal; any one reading is a sample, and should be reported as one.
Want to put this into practice?
Reachium automates LinkedIn outreach, content publishing, and inbox management in one platform.
See Reachium →Sources
- MarketScale, State of GEO: AI Visibility for B2B Companies. Reporting window August 4 to November 2, 2026; snapshot updated November 2, 2026; read 2026-11-03 via the MarketScale State of GEO study page. 169 projects, five engines, 11,934 prompts, 552,661 responses. Figures used: unweighted average mention rate 24.84%, average citation rate 17.98%, Google AI Mode mention rate 20.9% and citation rate 5.8%. Vendor-study figures, labelled as such throughout this post.
- Internal posts cited: how to measure AI search visibility (fixed prompt set measurement method), track AI citations (ongoing citation logging workflow). Checked 2026-11-04.
- The position that mention and citation rates must be tracked per engine rather than blended is the site operator's confirmed position, recorded 2026-11-03.
