Home/Blog/How to Rank in ChatGPT in 2026

Tactical · 8 minute read

How to get cited in ChatGPT in 2026.

ChatGPT does not rank pages. It retrieves them, reads them, and names some of them in a generated answer. This guide covers how that retrieval works, what OpenAI documents about its crawlers, what the published studies show, and the six things a B2B brand can do about it.

Key takeaways

Six things to know before you change anything.

How it works

How ChatGPT chooses what to cite.

When a ChatGPT answer includes web sources, three things have happened in sequence. A search index returned candidate pages for the question and for the sub-questions the model generated. A fetcher retrieved those pages. The model then wrote an answer and attached links to the passages it used.

The search step is the part most brands miss. OpenAI announced Bing as a search partner when ChatGPT search launched, and it runs its own crawler, OAI-SearchBot, to build an index alongside it. Seer Interactive's study is the clearest public measurement of the result: over 500 citations across 100 queries, with 87% of ChatGPT's citations matching Bing's top results and a median Google rank of 17 for the same URLs. If a page is not doing well in Bing, it has a hard time being retrieved by ChatGPT at all.

OpenAI documents three separate user agents, and they are not interchangeable. Blocking the wrong one costs nothing; blocking the right one removes you from answers.

OpenAI's crawlers as documented on developers.openai.com, read September 2026.
User agentWhat OpenAI says it doesHonors robots.txtAffects ChatGPT search
OAI-SearchBotSurfaces websites in search results in ChatGPT's search featuresYesYes. Opted-out sites are not shown in answers
GPTBotCrawls content that may be used in training foundation modelsYesNo
ChatGPT-UserVisits a page when a user asks ChatGPT a questionMay not apply (user-initiated)No. Not used to decide what appears in search

OpenAI says changes to OAI-SearchBot access take about 24 hours to process. The same page also notes that a site opted out of search may still appear as a navigation link when a user asks for it by name.

Step 1

Let the retrieval bots in, and give them HTML.

Check robots.txt for an explicit or wildcard block on OAI-SearchBot. Then check the CDN or firewall, which can block AI agents without touching robots.txt. Cloudflare has blocked AI crawlers by default on new zones since July 2025, so a site behind it may be opted out without anyone deciding to be.

Then look at what the bot receives. Vercel and MERJ analyzed hundreds of millions of requests across their network and found that OpenAI's, Anthropic's, Meta's, ByteDance's, and Perplexity's crawlers fetch HTML but do not execute JavaScript; only Google's Gemini (which uses Googlebot infrastructure) and Applebot render. The same study found ChatGPT's crawler spent 34.82% of its fetches on 404 pages and another 14.36% following redirects. A static or server-rendered page with clean internal links is the whole fix.

Reneka's own site is static HTML on nginx for exactly this reason: every passage an engine might quote is in the first response.

Step 2

Get indexed in Bing, and tell Bing when pages change.

Because of the Bing dependency, Bing Webmaster Tools is not optional for ChatGPT visibility. Verify the site, submit the XML sitemap, and read the Page Indexing report the way you would in Google Search Console. Bing's AI Performance report, in public preview since February 2026, then shows which pages Copilot cites and the grounding queries behind them, which is the closest thing to a first-party ChatGPT citation feed that exists.

Add IndexNow on top. It is a one-line POST that tells Bing and other participating engines a URL changed, so a refreshed page is recrawled in hours rather than weeks.

Step 3

Write passages a model can lift whole.

The only peer-reviewed test of on-page tactics for generative engines is Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024. The authors built a benchmark of 10,000 queries and measured how edits to a source page changed its share of the generated answer. Their abstract states the headline result:

GEO can boost visibility by up to 40% in generative engine responses.

The methods that worked were adding sourced statistics, adding attributed quotations, citing sources inline, improving fluency, and using precise technical vocabulary. The method that hurt was keyword stuffing. The paper also found the gains vary by topic, so the right edit for a legal explainer is not the right edit for a product comparison.

In practice that means each section of a page should be a complete unit: a heading phrased as the question, a direct answer in the first sentence, a number with its source, and no dependence on a paragraph above it. Put the answer before the context. Keep tables as real HTML tables and lists as real lists. Keep nothing important inside collapsed tabs or images.

Do not expect markup to do this work. Ahrefs tracked 1,885 pages that added JSON-LD against 4,000 controls between August 2025 and March 2026 and concluded that adding schema produced no major uplift in citations on any platform. Schema is worth keeping for Google and Bing rich results and for entity consistency. It is not a citation lever.

Step 4

Get named on the pages ChatGPT already cites.

Ahrefs analyzed 26,283 source URLs behind 750 top-of-funnel prompts and found that 43.8% of the pages ChatGPT cited were "best X" lists, and that 79.1% of the lists it cited had been updated within the year. For an agency, that means the roundups and comparison articles in the category are the citation surface. Getting listed on them, accurately and recently, matters more than almost anything on your own site.

The broader pattern is that engines follow mentions, not links. In the 75,000-brand Ahrefs correlation study, brand web mentions had the strongest relationship with AI Overview visibility at 0.664, branded anchor text came next at 0.527, and raw backlinks trailed at 0.218. The work is digital PR, guest contributions, podcast appearances with published show notes, and client case studies on the client's own site: anything that puts the brand name in a sentence on a page an engine trusts.

Step 5

Refresh pages for real, and show the date.

Ahrefs compared the publish and update dates of 16.975 million URLs cited by ChatGPT, Perplexity, Gemini, Copilot, and AI Overviews with organic Google results. AI-cited pages averaged 1,064 days old versus 1,432 for organic, a 25.7% difference, and ChatGPT's citations were 458 days newer than organic results on average. Google's AI Overviews showed almost no preference.

The tactic is a real refresh: new numbers, new examples, removed claims that stopped being true, then a visible updated date that matches the dateModified in the page's schema. Changing the date alone is a trust problem, and Google's own guidance calls it out.

Step 6

Measure citations and mentions as two numbers.

The Semrush and Growth Memo study of 3,981 domain appearances across 115 prompts, 14 countries, and four engines found that 61.7% of citations were ghost citations: the site was linked but never named. The engines also behave in opposite ways. ChatGPT cited a source in 87% of cases but named the brand in 20.7%. Gemini named brands 83.7% of the time and cited only 21.4%.

So a report that says "cited in 40% of answers" is telling you very little about whether a buyer ever saw your name. Track both. Run the same prompt set on a schedule, per engine, and record whether the brand was linked, whether it was named, and what else was named beside it. Reneka publishes how it does this on the measurement methodology page.

Where Reneka fits

What we do with this.

Every Reneka engagement runs the steps above as one program: crawler access and Bing indexing in the technical audit, passage-level rewriting in the content work, mention-building through digital PR, and a monthly prompt panel that reports citations and mentions per engine. The platform page describes the scorecard, and the free citation audit shows what ChatGPT, Perplexity, Gemini, and Claude say about your brand today.

FAQ

Common questions.

Does ChatGPT use Bing?

OpenAI named Bing as a search partner when ChatGPT search launched and also runs its own crawler, OAI-SearchBot. Seer Interactive's measurement found 87% of ChatGPT's citations matched Bing's top results, so Bing indexing is a practical prerequisite for being cited.

Which crawler do I need to allow to appear in ChatGPT?

OAI-SearchBot. OpenAI states that sites opted out of OAI-SearchBot are not shown in ChatGPT search answers. GPTBot only collects training data and ChatGPT-User only fetches pages a user asked about; neither decides search inclusion.

Does schema markup help get cited in ChatGPT?

Not measurably. Ahrefs compared 1,885 pages that added JSON-LD with 4,000 control pages over eight months and found no meaningful uplift on any platform. Schema remains useful for Google and Bing rich results and for keeping entity facts consistent.

How long does it take to get cited in ChatGPT?

Access changes take about 24 hours on OpenAI's side, and a refreshed page can be recrawled by Bing within hours if IndexNow is used. Earning mentions on the third-party lists ChatGPT already cites takes longer, typically weeks to months of outreach.

Get cited.
Not just ranked.

A 30-minute call with a Houston-based GEO strategist. We will run a live AI citation audit on your brand and walk you through what ChatGPT, Perplexity, Gemini, and Claude are saying about you today. No pressure, no pitch deck.