How should I write a blog post to get cited in AI search?

There is no AI writing style that earns citations. Write a clear post that answers one question well, put the key facts in plain sentences, and serve that text in the raw HTML.

The reasoning comes from how AI answers get built. An engine first decides whether it can fetch your page (access). Then it decides whether your page enters the model's context for a given query (retrieval). Only then does the model pick which retrieved sources to use and cite (generation).

Almost every published "GEO technique" was tested at that last stage, on pages already sitting in the model's context. A critical survey of 45 GEO studies (arXiv, July 2026) concluded that already-retrieved content can change its citation, but no technique shows a stable, cross-platform effect on being discovered in the first place. So style tweaks act on the step that matters least, and the steps before it are where you lose most posts.

What that means for your next post: spend your effort on topic fit, quotable facts and crawlable text, not on phrasing tricks.

Google is direct about it. Its guide to generative AI features says you don't need to write in a specific way for generative AI search. It also says you don't need to worry about covering every long-tail variation, and that there is no ideal page length. It rejects chunking content into blocks for LLMs as a requirement. Our guide on whether an AI-friendly style exists covers each claim in detail.

The controlled studies point the same way:

• C-SEO Bench (NeurIPS 2025) found most conversational SEO methods "not only largely ineffective but also frequently have a negative impact on document ranking."
• The 252,000-trial factorial study (arXiv 2605.25517) found formatting-only edits have little impact.
• The original GEO paper (KDD 2024) found keyword stuffing lowered its own visibility metric (17.8 against a 19.5 baseline).
• The 2026 survey warns that citation-oriented rewrites can impair retrieval.

The often-quoted "up to 40% more visibility" figure comes from a lab engine the authors built, with sources loaded into context. It says nothing about real discoverability. Research on predictable prose and AI citation found engines prefer text that is predictable to the model, but its causal test rewrote whole pages, so it isn't evidence for "write simply."

1. Topical fit to the real question

The 2026 survey and the 252,000-trial study both name topical relevance and context position as the biggest drivers of being cited. You can't control position. You can control fit. Write one post per question, answer it in the first lines, and cover what a buyer would need next. This is topical fit, not lexical matching, so skip the keyword lists.

2. Specific facts in plain sentences

A sentence like "The Starter plan costs $49 a month" is something an engine can lift and quote. Vague claims give it nothing. The factorial study found that including explicit price information helped consistently, though it ran on a synthetic corpus with content already in context. Put the fact in body text, not only in markup: see structured data vs plain-text facts. If you have numbers nobody else publishes, read whether original data gets cited.

3. Text in the HTML before JavaScript runs

Vercel's crawler study found ChatGPT and Claude crawlers fetch JavaScript files but don't execute them, so they can't read client-side rendered content. Googlebot does render. If your blog loads post bodies, tables or pricing through client-side scripts, those parts may be invisible to most AI crawlers. Server-render the content that matters.

Good writing can't fix problems upstream of the page. Check these before you rewrite anything.

• The crawler is blocked. robots.txt rules are per bot. OAI-SearchBot controls ChatGPT search, while GPTBot only governs training. A firewall or CDN can also return a 403 to a bot that robots.txt allows. The AI Crawler Checker requests a page as each AI crawler and shows which ones get turned away.
• The page isn't eligible on Google. Google says a page must be indexed and eligible for a snippet to appear as a supporting link in AI Overviews or AI Mode. noindex, nosnippet and max-snippet:0 keep content out.
• The engine didn't search. Retrieval is per query and discretionary. An answer given without live search draws only on what the model learned before its cutoff, and nothing you publish this quarter reaches it.
• The engine cites other sites. When answers lean on reviews, directories or comparison posts, being named on those sites is the route in. That is an association in the evidence, not a promised effect.
• The wrong page gets picked. If a post wins over the page you wanted, see why AI cites a blog post instead of a product page.

Volume is not a workaround. In the Otterly 2,000 AI blogs experiment, Google deindexed both mass-published sites, and its AI surfaces almost never cited them after that.

1. Pick one buyer question and answer it in the first two sentences.
2. State the facts a buyer needs (prices, limits, dates, definitions) as plain sentences.
3. Keep those facts consistent with your pricing and about pages.
4. Load the page with JavaScript disabled and confirm the body text is there.
5. Confirm robots.txt allows OAI-SearchBot, PerplexityBot and Claude-SearchBot, and that your CDN doesn't block them.
6. Check index status in Google Search Console and Bing Webmaster Tools separately.
7. After publishing, ask the engines your question. The free checker on this site shows whether AI names you for up to three buyer questions, as a one-time snapshot.

You control access, rendering and what your page says. You don't control what the models choose to cite. Platform behavior changes, so treat this as the state of the evidence as of October 2026.