How to earn more citations from AI assistants

First, check assistants can read your site

None of the later steps count if an assistant can't read the page. Vercel and MERJ looked at how AI crawlers behave and found that GPTBot, OpenAI's crawler, fetches JavaScript files but doesn't execute them. If your content only appears once scripts have run, that crawler doesn't see it. These three checks take very little time:

1. Turn JavaScript off in your browser and reload your key pages. The words you want cited should still be there.
2. Read your robots.txt. Unless you've chosen to opt out, it should allow GPTBot, ClaudeBot, PerplexityBot and Google-Extended.
3. Look beyond Google. ChatGPT and Copilot retrieve through Bing, and Claude through Brave, so a page missing from those indexes can miss out however well it does in Google.

Get mentioned on other people's pages

This is usually where the biggest gains are. When Ahrefs studied 75,000 brands, web mentions correlated with visibility in Google's AI Overviews at 0.664, against 0.218 for backlinks. That's a correlation in one product, not proof of cause, but it fits the wider picture. AirOps found that 85% of brand mentions come from third-party pages rather than a brand's own domain, and that brands are 6.5 times more likely to be cited through third-party sources than their own. Academic work points the same way: Chen and colleagues describe a strong AI search bias towards earned media.

In rough order of priority:

• Other people's "best of" lists. Third-party lists take 80.9% of listicle citations, and self-promotional ones just 19.1%. Find the lists your assistants already cite and make a real case for being added.
• Review sites. Keep your profiles complete and ask real customers for reviews on the platforms your market uses.
• Press and trade coverage. Give journalists something worth quoting, such as your own data or a clear expert view.
• Communities and video. Take part honestly in forums where your buyers talk, and say who you are. Fake accounts aren't worth the risk.

Your own site still matters

Off-site work doesn't replace your own pages. In Pratyush Kumar's study of more than 100,000 prompt responses, about 78% of citations went to corporate websites. Which kind of page you build matters too. The same AirOps page type research puts listicles at 21.9% of citations, articles at 16.7% and product pages at 13.7%.

Let the prompt decide the page. If someone wants to learn, write an article. If they're comparing, publish a fair comparison. If they're close to buying, put full details on the product page itself, including specifications and what's included, rather than hiding them in a brochure.

Write pages that are easy to cite

The original GEO study from Princeton and others measured visibility gains of up to 40% from changes such as adding statistics, quotations and sources. A 2026 review of 45 studies adds an important limit: those gains depend on your page already being in the assistant's context. They help you get quoted once you've been found, not found in the first place. With that in mind, these habits are worth keeping:

• Start with the direct reply. Say the main thing in the first lines, with nothing above it.
• Back up claims. Give the number, say where it came from and when, and link to the original.
• Use one idea per section, under a heading that says what the section covers.
• Say who wrote it, and why they're worth listening to.
• Publish something only you have. Your own data or tests give others a reason to mention you, and rivals can't copy them.

Cover the follow-ups

Assistants often turn one prompt into several searches. In Surfer's study of 10,000 keywords, pages ranking for both the main query and its fan-out queries were 161% more likely to be cited in Google's AI Overviews than pages ranking for the main query alone. In practice, a page on choosing a product should also deal with what people ask next, such as size, cost, alternatives and common problems.

Keep pages current

Ahrefs analysed about 17 million citations and found that content cited by AI assistants was 25.7% fresher than the organic search results it was compared with. AirOps adds that pages not updated quarterly are 3 times more likely to lose citations. So put your most important pages on a quarterly review, and make real changes when you do: new figures, new examples, corrected details. Changing the date alone adds nothing for a reader.

What not to expect

Be wary of anyone selling a formatting trick as a guarantee. The peer-reviewed C-SEO Bench study tested these methods directly and reached a blunt conclusion:

Most current C-SEO methods are not only largely ineffective but also frequently have a negative impact on document ranking.

C-SEO Bench, NeurIPS 2025

The authors also found that traditional SEO did better, and that gains shrank as more competitors adopted the same tactic. A trick that works because few people use it stops working once everyone does. Being well known, being mentioned by others and having facts nobody else has are much harder to copy.

Results also take time to read, because they vary from run to run. Before you change anything, write a steady set of prompts and note where you stand, then check the trend after each change. Suggestible runs those prompts for you, shows which pages get cited for you and your rivals, and tells you what to fix first.