Which AI engine is most likely to cite you? What 48 audits show

In 773 measured AI answers across five engines, brands were cited 48 times, a 6.2% overall citation rate. The spread between engines is the real finding: Perplexity cited a brand in 13.4% of measured answers, Google AI Overviews in 3.9%, Gemini in 2.2%, Claude in 1.5%, and ChatGPT in 1.2%. If your GEO strategy treats "AI visibility" as one number, it is averaging away a 10x difference.

Disclosure: these are BeCited's own audit measurements, published as field notes. The dataset is small (48 audits, 9 domains, measured between May 19 and August 24, 2026) and skews toward B2B software sites, including our own. We publish the counts, not just the percentages, so you can judge the weight of the evidence yourself.

What did we measure?

Every BeCited audit asks the same set of buyer queries to each AI engine and records, per query and per engine, whether the answer cites the audited domain. The method is deterministic: the same audit re-run gives the same score, which is what makes the numbers below comparable across months (how the CITE score works).

The dataset behind this post:

• 48 audits over 9 domains, May 19 to August 24, 2026
• 1,132 engine-query cells, of which 773 produced a measurable answer
• 48 cells where the answer cited the audited brand

Which engine cites brands most often?

Perplexity, by a wide margin. In our measurements Perplexity cited the audited brand in 36 of 269 answers (13.4%). Google AI Overviews followed at 6 of 154 (3.9%), Gemini at 3 of 135 (2.2%), Claude at 2 of 132 (1.5%), and ChatGPT at 1 of 83 (1.2%).

Two things stand out.

First, the ordering matches the largest public study we know of: a 2026 analysis of 34,000 AI answers put ChatGPT's citation rate near 0.6% and Perplexity's near 13%, a 46x spread at the extremes. Our sample is a fraction of that size, but it reproduces the same shape from independent measurements, which is what you want from two datasets describing the same reality.

Second, the engines that cite least are the engines with the most usage. A strategy that optimizes only for ChatGPT is optimizing the engine least likely to send a citation your way; a strategy that ignores Perplexity is ignoring the engine that pays out most often.

Why do the rates differ this much?

Because the engines are built differently. Perplexity is retrieval-first: it searches, then writes from sources, so citations are structural. ChatGPT and Claude answer from model knowledge for many queries and only sometimes browse, so most answers carry no citation at all. Google AI Overviews cite, but from a set Google's systems select, which overlaps surprisingly little with the other engines.

This is also why measuring a single engine misleads. In our data, the domains that earned Perplexity citations mostly earned nothing on ChatGPT for the same queries. Public research points the same way: reported domain overlap between ChatGPT and Perplexity citations is around 11%. Being cited on one engine is weak evidence about any other.

What should a small brand do with this?

• Measure per engine, not in aggregate. An average across engines hides where you are actually winning (why brands go missing on ChatGPT).
• Treat Perplexity as the fastest feedback loop. It cites an order of magnitude more often, so fixes show up there first; use it as the canary for your GEO work.
• Expect low absolute numbers. A 6.2% overall citation rate means that for most buyer queries, today, nobody in your category is being cited either. The gap is open.
• Re-measure with the same method. Rates this low are noisy run to run unless the measurement is repeatable; deltas only mean something when the method holds still.

We will keep publishing these field notes as the dataset grows. The counts above update with every audit that runs, and the free report measures your own domain against the same engines in about a minute.