How to get cited by ChatGPT: the 9 signals that actually matter

"How to get cited by ChatGPT" grew roughly 933% quarter over quarter in US search, per Google Ads data pulled in September 2026. Nobody can guarantee you a citation, and anyone promising one is worth distrusting. But across a large number of observed ecommerce answers, patterns repeat consistently enough to work from. Here are nine signals, ordered by return on effort.

The three to fix first

These take days, not quarters, and they gate everything else. Until they are resolved, the rest of the work produces nothing measurable.

• Crawler access: GPTBot and OAI-SearchBot need to be able to read your pages. An inherited block from an old configuration is the single most common problem we find.
• Readability without JavaScript: if product content only exists after scripts run, part of the crawler population never sees it.
• Accuracy of the basics: price, availability and product name consistent across your page, your markup and your feed.

The three that decide the citation

• Factual density: a page stating material, dimensions, compatibility and care gives the engine something it can restate without risking an error.
• Match to the real question: people ask about a use case, not a category. A page that names the use case has a direct advantage.
• Disambiguation: if your brand name collides with a common word, the engine needs help resolving it, and clean markup provides that.

The three slow ones, and why they weigh most

• External corroboration: a fact confirmed on a third-party site is restated with far more confidence than one that exists only on yours.
• Presence in category roundups: these are preferred sources when the question asks for a choice between brands.
• Consistency over time: a brand described identically everywhere, for a long time, becomes the engine's default safe answer.

What does not work

Two bad ideas circulate widely. Writing instructions aimed at the model into your page, visible or hidden, is treated as manipulation and can count against you. Mass-publishing generated content to saturate a topic raises volume, not perceived reliability, and perceived reliability is what decides.

The method that works is duller. Identify the questions where you lose, look at what attribute the brand cited in your place actually has, add that attribute to your own record, then check whether the answer moved. That is the loop Tadow automates, with a before and an after, and an explicit verdict when the gap is too small to credit to your fix.