How to Get Cited by ChatGPT: The Three Signals That Actually Correlate

The top three signals are all off your own site

Ahrefs studied 75,000 brands. The result is not what most optimisation checklists assume.

They measured how strongly each factor correlates with a brand appearing in AI Overviews, using Spearman correlation across domains above DR 40. Here is the order.

Factor	Correlation with AI visibility
Branded web mentions	0.664
Branded anchor texts	0.527
Branded search volume	0.392
Domain Rating	0.326
Number of backlinks	0.218

Read the top three rows again. Mentions, anchors and search volume are all things other people do about your brand, not things you edit in your theme.

Still, that is a genuinely awkward finding for a checklist-shaped industry. It means the highest-leverage work is getting talked about, which is slower and less tidy than editing a meta field.

YouTube is the strongest single signal

A follow-up study widened the lens. Specifically, it covered ChatGPT, Google AI Mode and AI Overviews together.

Across all three, YouTube mentions correlated at roughly 0.737 — higher than every other factor measured, including branded web mentions overall. So being discussed on YouTube is, statistically, the closest thing to a universal AI visibility signal currently identified.

In fact, that squares with where citations actually land. YouTube is the single most-cited domain in Google AI Overviews at 21.1%, with Reddit at 18.5% and Facebook at 10.7%.

For a Shopify merchant that does not mean launching a channel tomorrow. It means unboxings, reviews, comparisons and demos featuring your product are doing work you cannot do from your admin panel.

Your own product pages still matter, for a different reason

Off-site signals win on correlation. But owned pages win on share of citation.

In AI shopping answers, brand-owned and manufacturer sites take 64.7% of citations, against 2.9% for retailers and marketplaces. That is roughly a 22 to 1 split, measured across 391,073 citations from around 1,468 non-branded shopping queries.

So the two findings are not in tension. Off-site mentions make an assistant aware of you, and your own product pages are what it quotes once it is. We covered that split in detail in why AI assistants cite brand sites over retailers.

So the practical consequence is unglamorous. A product page with two lines of copy gives an assistant nothing to lift. So thin catalogues lose citations they were already in line for.

What the data does not support: schema as the lever

This one deserves care, because a lot of GEO marketing leans on it. Schema markup does not appear among the top measured correlating factors in any of the studies above.

That is not a reason to skip it. Schema is parsing hygiene. It structures your product data so machines read price, availability and reviews correctly. Microsoft has also said schema helps Copilot understand page content. It is cheap, it has value well beyond AI search, and you should have it.

But treat it as hygiene rather than as the lever. No study currently ranks it among the signals that move citation rates. So a store with perfect markup and no mentions stays largely invisible. Our guide to schema markup and AI search covers the on-site side properly.

Freshness helps, but less than you would think

Recency is real here. It is also frequently oversold.

Across 16.975 million cited URLs on seven platforms, AI-cited content averaged 1,064 days old against 1,432 days for organic top-ten results. That is a 25.7% freshness advantage, not a multiplier, and the average cited page is still nearly three years old.

So keep publish and update dates visible and refresh your important pages. Just do not expect a date stamp to carry a page that nobody mentions.

The honest caveat

Every figure above is a correlation. Ahrefs says so directly, and it matters.

None of this proves that earning mentions causes citations. Brands that get talked about a lot tend to be bigger, better known and more linked, and those things travel together. Anyone selling you a guaranteed citation formula is ahead of the evidence.

It also varies by platform. Only about 11% of domains cited by ChatGPT are also cited by Perplexity, which we broke down in the ChatGPT and Perplexity citation gap.

A checklist you can actually act on

Ordered by what the evidence supports, not by what is easiest.

1. Get your products into the hands of people who make video. YouTube mentions are the strongest single correlating signal found so far.
2. Earn unlinked brand mentions in communities, roundups and press. Mentions beat backlinks by roughly three to one here.
3. Give every product page enough substance to quote. Thin pages cannot be cited even when the assistant reaches them.
4. Add schema because it is cheap and correct, but do not count it as your AI strategy.
5. Show publish and update dates, and refresh your most important pages on a real schedule.
6. Check more than one assistant, because their source pools barely overlap.

Which guide you need

Read this one for the off-site picture. The others go deeper on their own piece.

For the on-site technical layer, use the schema markup guide. For which tools measure any of this, see the best Shopify apps for AI search visibility.