Is there any proven way to increase AI citations?

The most thorough review so far says no. A critical survey of 45 GEO studies, posted to arXiv in July 2026, concluded that already-retrieved content can causally alter its citation or use. It also found that no reviewed technique shows a stable, longitudinal, cross-platform causal effect on organic discoverability.

That splits the problem in two. An AI answer passes through four stages:

• Parametric: what the model knows without web access, frozen at a training cutoff. Nobody changes this quickly.
• Access: whether a crawler may fetch and store your page. You and your CDN control this.
• Retrieval: whether your page enters the model's context for a question. The engine controls this, partly.
• Generation: which retrieved pages the model uses and cites. The model controls this.

Almost every GEO experiment tests the last stage, on pages already in context. Most GEO advice promises the earlier ones. So you can't buy a citation. You can stop blocking one.

The well-known numbers come from lab setups, not live search.

• The "up to 40%" GEO figure comes from the 2023 GEO paper. The authors built their own engine and loaded the sources into its context. Even their Perplexity test uploaded source text as files.
• C-SEO Bench (NeurIPS 2025) found most conversational SEO methods largely ineffective and frequently harmful to a document's ranking. Gains also shrank as more sites adopted them.
• The 252,000-trial factorial found formatting-only edits had little impact. Topical relevance and list position drove which source got cited first. It never called a live search engine and used two candidate documents per trial.
• Ahrefs' schema test tracked 1,885 pages that added schema against matched controls. It found no major uplift on any platform, and AI Overviews citations fell 4.6%. Every treated page already had 100+ AI Overview citations, so whether schema helps a page enter the candidate set is untested.

What this means: content rewrites can shift citation among pages already retrieved. Nothing shows they get you retrieved.

The access levers work by subtraction. Their mechanisms are documented by the engines themselves.

• Server-render the content that matters. Vercel found ChatGPT and Claude crawlers fetch JavaScript files but don't execute them, so they can't read client-side rendered content. Googlebot renders JavaScript, so this matters most outside Google.
• Allow the right bot. OpenAI states GPTBot governs training and OAI-SearchBot governs ChatGPT search, and that each setting is independent. Sites opted out of OAI-SearchBot won't be shown in ChatGPT search answers. Check PerplexityBot and Claude-SearchBot the same way.
• Check the firewall. A CDN or WAF can return 403 to a crawler that robots.txt allows. From 2026-09-15, Cloudflare blocks Training and Agent crawlers by default on ad-displaying pages for new domains and free-tier customers, so access can change without you editing anything.
• Remove accidental suppression. noindex, nosnippet, max-snippet:0 and data-nosnippet keep content out of AI answers. Templates often carry them by mistake.
• Stay eligible in Google. Google says a page must be indexed and snippet-eligible to appear in AI Overviews or AI Mode, with no additional requirements. More in how to get cited in AI Overviews and AI Mode.

The AI Crawler Checker requests a page as each AI crawler and shows which ones your firewall turns away.

State specific facts on fetchable pages

A price, a number or a one-line definition in plain HTML gives the engine a sentence it can quote. This is the only positive experimental signal for on-page content. The 252,000-trial factorial found explicit price information and a recent timestamp helped consistently, in a lab setup.

Say you run a project-management SaaS and your plan prices load in a JavaScript widget. Put those prices in the HTML, stated the same way on every page.

Get named where engines already cite

When engines cite other sites instead of yours, being named on those sites is the route into the answer. The evidence shows a strong association, not a promised effect. Brand size confounds every study here. Engines also draw on different sources, so check whether what works on one engine works on the others.

These are sold widely. Primary sources contradict them as citation levers.

• Schema or FAQ schema to get cited. Google says structured data isn't required for generative AI search, and FAQ rich results were deprecated in May 2026. Existing markup does no harm. It just isn't a lever.
• llms.txt. Google says Search ignores it. No engine documents reading it.
• E-E-A-T scores and domain authority. Google says E-E-A-T isn't a specific ranking factor. Domain Authority is a Moz metric no engine computes.
• Chunking or AI-friendly writing. Google rejects both and says there is no ideal page length.
• Allowing GPTBot for visibility. It governs training only.
• IndexNow pings and AI rank tracking. The IndexNow spec never mentions AI, and Google exposes AI impressions only, with no position data.

The full sorted list is in AI SEO best practices, sorted by evidence.

1. Diff your key pages with and without JavaScript. Anything that appears only after rendering is invisible to most AI crawlers.
2. Read your robots.txt bot by bot. Confirm OAI-SearchBot, PerplexityBot and Claude-SearchBot are allowed if you want search visibility.
3. Request a page as each crawler and confirm your CDN returns 200, not 403 or a challenge.
4. Search templates for stray noindex, nosnippet and data-nosnippet.
5. Put prices, numbers and definitions in plain HTML, consistent across pages.
6. Find the sites engines cite for your buyer questions and work on being named there.

Then measure each engine separately. The free checker on this site shows whether engines name you for three buyer questions, as a snapshot. For sequencing, see GEO optimization: a practical order of work.