Why does AI cite my blog post instead of my product page?

AI search engines don't search for your question as typed. As of October 2026, OpenAI says ChatGPT search typically rewrites a prompt into one or more targeted queries. Google describes a similar query fan-out for AI Mode. Each of those queries pulls back candidate pages, and the model cites from that set.

Most buyer questions are informational: "how does X work", "what does X cost", "X vs Y". A blog post that answers that question in its first paragraph is a strong topical match. A product page built around a hero image, a feature grid and a "book a demo" button often contains no sentence that answers anything.

Research backs the pattern. A 2026 critical survey of 45 GEO studies found topical relevance and position in the model's context are the most reproducible drivers of citation. The 252,000-trial factorial reached the same conclusion. You can influence relevance. You can't see or set position.

So the blog citation is often the engine doing its job. The question is whether your product page was ever a real candidate.

Product pages are the pages most likely to be built as JavaScript applications. That matters outside Google.

Vercel's crawler study found that ChatGPT and Claude crawlers fetch JavaScript files but don't execute them, so they can't read client-side rendered content. The finding was independently re-confirmed in June 2026 for OpenAI, Anthropic, Perplexity, Meta and ByteDance crawlers. Googlebot and AppleBot are the exceptions: they render. Read the Vercel study digest for its limits.

Common ways a product page goes blank for an AI crawler:

• Pricing loaded from an API after the page mounts
• Specs inside tabs or accordions that fetch content on click
• Key claims set in an image or a video, with no text equivalent
• Comparison tables rendered by a script

Meanwhile, a blog post on a static CMS usually ships its full text in the first HTML response. The crawler sees one page with words and one without.

Stating a specific fact (a price, a spec, a one-line definition) in fetchable HTML gives the engine a sentence it can quote. It's the only positive on-page signal in the evidence. Engines quote at different grain, as covered in whether AI cites whole pages or sentences, but none can quote text it never received.

A product page can also be missing for reasons that have nothing to do with its words.

Per-bot robots.txt rules

Rules apply bot by bot. OpenAI states that each of its crawler settings is independent. GPTBot controls training use. OAI-SearchBot controls ChatGPT search. A product directory blocked for one bot isn't blocked for the others, and the reverse is true too.

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

That example allows ChatGPT search and opts out of training. Check that no rule blocks /product/ or /pricing/ for OAI-SearchBot, PerplexityBot or Claude-SearchBot.

Firewall and CDN responses

OpenAI's help centre says eligibility requires allowing OAI-SearchBot and confirming your host or CDN accepts its published IP addresses. A firewall can return 403 to a bot that robots.txt allows. Product and checkout paths often sit behind stricter bot rules than the blog does.

Snippet controls on templates

Google and Bing document noindex, nosnippet, max-snippet:0 and data-nosnippet as the controls that keep content out of AI answers. Product templates sometimes inherit these by accident. On Google, a page must be indexed and eligible for a snippet to appear as a supporting link in AI Overviews or AI Mode.

If crawlers hit your blog but not your shop, see why AI crawlers visit but never cite.

Be wary of fixes that promise to steer citations to a chosen URL.

• Schema. Google says structured data isn't required for generative AI search. Ahrefs' matched test of 1,885 pages found AI Overview citations fell 4.6% after schema was added. Existing markup does no harm, but it isn't a lever. See the Ahrefs schema study digest.
• Rewriting for AI. C-SEO Bench found most citation-oriented rewrites ineffective and often harmful.
• Hidden text or cloaking for AI bots. Content shown only to crawlers is manipulation. Research shows injected content can sway some engines, which is exactly why it's a risk: it can be detected, and it breaks trust when it is.

Also remember a citation is not the whole story. OpenAI exposes retrieved sources separately from citations, and the retrieved set is larger. Your product page may have been read and not cited. The guide to cited vs retrieved sources explains the gap.

Work through these in order. Each one removes a block; none guarantees a citation.

1. Diff raw HTML against the rendered page. Fetch a product page with JavaScript off. If the price, core specs or one-line description are missing, move them into the server response.
2. Put facts in text. Write the price, plan limits and the plain-English "what it is" as HTML sentences, not images.
3. Test each crawler. Request the product URL as OAI-SearchBot, PerplexityBot and Claude-SearchBot. The AI Crawler Checker shows which ones the firewall turns away.
4. Audit template tags. Search product templates for noindex, nosnippet and data-nosnippet you didn't intend.
5. Link the blog post to the product page. Readers who arrive from a citation should reach the product in one click.
6. Measure per engine. Use the free AI citation checker to see which URLs each engine cites for your buyer questions, then recheck after fixes ship.