How to Rank in ChatGPT Search: A 2026 Playbook

You rank in ChatGPT search by being the page ChatGPT picks when it runs its own follow-up searches: let OpenAI’s search crawler in, get indexed in Bing, write pages that answer the narrow sub-questions behind a prompt, and earn mentions on the sites it already trusts. That is the short version of how to rank in ChatGPT search. The rest of this guide shows the exact robots.txt lines, the Bing and IndexNow setup, the page patterns that get cited, how to measure it in GA4, and what it costs a small team to do properly.

One warning before the steps. Most “ChatGPT ranking factor” lists online are vendor opinions. This guide sticks to what OpenAI documents and what large public studies measured, and it says so when a recommendation is a judgment call rather than a finding.

How ChatGPT search picks the pages it cites

ChatGPT search does not show ten blue links. It reads the prompt, writes several of its own search queries, retrieves pages, and then cites a handful of them in one answer. That middle step is called query fan-out, and it is the single most useful idea for anyone working on ChatGPT visibility.

The best public measurement of this comes from an AirOps study published in March 2026. It started with 15,000 prompts, which ChatGPT expanded into 43,233 queries, and tracked 548,534 retrieved pages. Only about 15% of retrieved pages were cited in the final answer. ChatGPT generated two or more fan-out queries on 89.6% of searches, and 95% of those fan-out queries had zero monthly search volume in keyword tools. In plain terms: the queries that decide whether you get cited are ones you will never see in a keyword tool.

The retrieval layer is not OpenAI’s alone. Since launch, ChatGPT search has drawn on third-party search providers, and Yoast’s analysis of the launch describes Microsoft Bing as central to how it works. OpenAI also runs its own crawler, OAI-SearchBot, which its bot documentation says is used “to surface websites in search results in ChatGPT’s search features.”

Traditional rankings still help, but less than you might expect. AirOps found that pages ranking first in Google were cited 43.2% of the time, 3.5 times the rate for pages outside the top 20. Yet an Ahrefs study of 15,000 long-tail prompts found only 8% of ChatGPT’s in-text citations ranked in Google’s top 10 for the original prompt. Both are true at once: ranking helps, but ChatGPT is ranking you for the fan-out queries, not for the prompt the user typed. Ahrefs explains that results across query variations are merged, so a page that ranks sixth for several variations can beat a page that ranks first for only one.

Who should invest in ChatGPT search, and who can wait

Not every business needs a ChatGPT program this quarter. The work pays off fastest when buyers research the category with long, specific questions before they talk to a vendor, because that is exactly the prompt type that triggers fan-out searches and citations.

Strong fit:

• B2B services and software where buyers compare options (“best CRM automation partner for a 20-person agency”). These prompts produce shortlists, and a citation puts you on the shortlist.
• Companies that already publish detailed how-to content. AirOps reported higher citation rates for product discovery and how-to style queries than for validation queries, so existing guides are an asset.
• E-commerce brands with clean product data, which can also use ChatGPT’s merchant feeds (covered later).

Weak fit for now:

• Purely local businesses whose customers search by map. Fix your Google Business Profile and Bing Places first.
• Sites with almost no indexed content. Without pages in the Bing and Google indexes, there is nothing for ChatGPT to retrieve. Build the content base first; our guide to AI-assisted keyword research is a reasonable starting point.
• Teams with no analytics hygiene. If you cannot see referral sources today, you will not be able to tell whether any of this worked.

If you are in the weak-fit group, still do Step 1 below. It takes ten minutes and prevents an accidental block that would cost you later.

Step 1: Make sure OAI-SearchBot can crawl your site

This is the only step OpenAI explicitly tells site owners to take. Its publishers and developers FAQ says to make sure you are not blocking OAI-SearchBot, and the bot documentation states that sites opted out of OAI-SearchBot “will not be shown in ChatGPT search answers,” though they can still appear as plain navigational links.

OpenAI runs separate crawlers for separate jobs, which is where most mistakes start:

• OAI-SearchBot surfaces sites in ChatGPT search. Allow it.
• GPTBot collects data for model training. Blocking it does not remove you from search; it is an independent setting.
• ChatGPT-User fetches a page when a user asks ChatGPT to visit it. OpenAI says it is not used to decide whether content appears in search.
• OAI-AdsBot only visits pages submitted as ads.

A robots.txt that keeps you in search while opting out of training looks like this:

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

User-agent: *
Disallow: /wp-admin/
Allow: /wp-admin/admin-ajax.php

Sitemap: https://example.com/sitemap_index.xml

Whether to block GPTBot is a business choice, not a ranking one. Many agencies allow both. Either way, check that a security layer is not blocking the bot even when robots.txt allows it. Firewalls and bot-protection services often challenge unfamiliar user agents. Test from the command line:

curl -s -o /dev/null -w "%{http_code}\n" -A "OAI-SearchBot/1.0; +https://openai.com/searchbot" https://example.com/
curl -s https://openai.com/searchbot.json | head -20

A 200 is good; a 403 or a challenge page means your firewall needs an allow rule. OpenAI publishes the crawler’s IP ranges at the searchbot.json address above, which is the safe way to whitelist it without trusting the user-agent string alone. After a robots.txt change, OpenAI says it “can take ~24 hours” for its systems to adjust, so do not panic if nothing changes the same afternoon.

To keep a specific page out of ChatGPT search, OpenAI’s FAQ points to the noindex meta tag, and notes the crawler must be allowed to fetch the page to see it. Blocking the URL in robots.txt is not the same thing.

Step 2: Get every important page into Bing, then keep it fresh with IndexNow

If ChatGPT retrieves through third-party search providers and Bing is one of them, a page Bing has not indexed is a page ChatGPT is less likely to find. Yoast puts it directly: getting your content indexed by Bing is crucial if you want it to be visible. Many small sites have never opened Bing Webmaster Tools because Bing sends them little direct traffic. That reasoning no longer holds.

1. Create a Bing Webmaster Tools account and import your site from Google Search Console. The import copies verification and sitemaps in a few clicks.
2. Submit your XML sitemap and check the indexing report for pages marked “Discovered but not crawled” or excluded.
3. Use URL Inspection on your five most important commercial pages. Fix anything that is not indexed before you write a single new article.
4. Turn on IndexNow so Bing hears about new and updated pages the moment you publish.

IndexNow is a simple protocol: you host a key file at your site root and send changed URLs to a participating search engine, which shares them with the others. On WordPress, Bing’s IndexNow page lists support in Rank Math, Yoast, AIOSEO, SEOPress and a dedicated IndexNow plugin, so most sites only need to flip a setting. If you prefer to do it by hand, or from an automation tool such as n8n, the request is one line:

# 1. Host the key: https://example.com/3f9c2a7e5b1d4c88.txt containing 3f9c2a7e5b1d4c88
# 2. Ping a single updated URL
curl "https://api.indexnow.org/indexnow?url=https://example.com/updated-page/&key=3f9c2a7e5b1d4c88"

# 3. Or send a batch (up to 10,000 URLs per request)
curl -X POST https://api.indexnow.org/indexnow \
  -H "Content-Type: application/json; charset=utf-8" \
  -d '{"host":"example.com","key":"3f9c2a7e5b1d4c88","urlList":["https://example.com/a/","https://example.com/b/"]}'

Bing notes that a submission does not guarantee crawling or indexing, and that you should only submit URLs that actually changed. Treat IndexNow as a doorbell, not a guarantee. Google rankings matter too, given the AirOps finding above, so keep your normal Google SEO going in parallel.

Step 3: Write pages that answer the fan-out queries

This is where most of the leverage is, and where the step from “retrieved” to “cited” happens. Remember that only 15% of retrieved pages made it into answers in the AirOps data. The same study found that pages whose title overlapped the query by 50% or more were cited 20.1% of the time, against 9.3% for pages with under 10% overlap. Titles that literally name the question win.

Work out the fan-out queries for your topic before writing. A practical method:

1. Write the buyer’s real prompt, long and specific: “We’re a 15-person accounting firm. What’s the cheapest way to automate client onboarding emails?”
2. List the sub-questions an assistant would need to answer it: tools that do this, their prices, setup steps, limits, alternatives, risks. Add the current year to research-type queries; AirOps saw year modifiers in nearly 10% of research fan-out queries.
3. Make each sub-question either an h2/h3 on one comprehensive page or its own focused page, depending on how much there is to say.

Then shape each section so it is easy to lift into an answer:

• Answer first. The first sentence under a heading should be the answer, not a scene-setter. An assistant extracting a passage takes what is at the top.
• Name things. Real tool names, plan names, prices and version numbers. Vague claims give the model nothing to cite.
• Use tables for comparisons. A table of options, prices and limits answers the “vs” and “best” sub-queries in one block.
• Keep it readable. AirOps reported that cited pages more often had Flesch Reading Ease scores of 50 or higher. Short sentences and plain words help both humans and extraction.
• Date and update. Show a visible “last updated” date and actually refresh prices and screenshots. A stale price is a reason for the model to cite a competitor.

Authority matters less than people assume. AirOps found that sites with domain authority between 40 and 80 were cited more often than the biggest sites, and the DA 80 to 100 tier had the lowest citation rate at 15.0%. A specialist page that answers the exact sub-question can beat a famous generalist. For a broader view of how this fits into AI search work, read our generative engine optimization guide.

Step 4: Earn mentions on the sources ChatGPT already trusts

ChatGPT builds its answer from several sources, so your brand can appear even when your own page is not cited: a review site, a comparison article, an industry newsletter or a forum thread may mention you. That is why off-site work belongs in any plan for ranking in ChatGPT search.

Be careful with the popular “just post on Reddit” advice. Reddit’s share of ChatGPT citations has swung wildly. Semrush reported that Reddit appeared in about 60% of ChatGPT responses in early August 2025 and roughly 10% by mid-September 2025, and that a separate tracker saw Reddit’s citation share drop from 3.8% to 0.5% in August 2026, a figure the tracker itself called provisional. A tactic that depends on one platform’s weighting can disappear overnight.

The durable version of off-site work looks like ordinary good PR, aimed at the pages that show up in your fan-out queries:

1. Run your 20 most important buyer prompts in ChatGPT and record which third-party URLs it cites.
2. Group them: listicles and “best X” roundups, review platforms, partner directories, community threads, news.
3. For roundups, pitch the author with something they can use: a current price table, a customer result, an expert comment.
4. Claim and complete your profiles on the review and partner directories that appear (G2, Clutch, tool partner pages).
5. Join community threads only where you can answer the question genuinely, with your name and role visible.

Keep your company description identical everywhere: same name, same one-line description of what you do, same category words. Consistent descriptions make it easier for a model to connect the mentions to one entity. Tracking those mentions is a job worth automating; our post on autonomous brand monitoring shows one way to set that up.

Step 5: Measure ChatGPT traffic and citations

There is no ChatGPT Search Console, so measurement takes three pieces: referral traffic, crawler logs and a fixed prompt panel.

Referral traffic. OpenAI’s publisher FAQ says ChatGPT automatically adds utm_source=chatgpt.com to referral URLs. In GA4, build an exploration filtered on session source, or create a custom channel group for AI assistants with a regex like this:

^(chatgpt\.com|chat\.openai\.com|perplexity\.ai|copilot\.microsoft\.com|gemini\.google\.com|claude\.ai)$

Put the AI channel above “Referral” in the channel order, or the sessions get absorbed into generic referrals.

Crawler logs. Check that OAI-SearchBot is actually visiting, and which pages it fetches. On a typical Nginx or Apache server:

grep "OAI-SearchBot" /var/log/nginx/access.log | awk '{print $7}' | sort | uniq -c | sort -rn | head -20

If your important pages never appear in that list, go back to Steps 1 and 2.

Prompt panel. Keep a list of 20 to 50 buyer prompts and run them on a fixed schedule, recording whether you are cited, mentioned or absent. You can do this by hand in a spreadsheet or with a tracker. Peec AI, for example, starts at $95 a month for 50 prompts across three engines you choose. Bing’s AI Performance report shows citations in Copilot and Bing AI answers for free, but Adobe’s guide notes it does not include ChatGPT, so treat it as a proxy for Bing-side retrieval rather than a ChatGPT report.

For turning these three sources into one weekly view instead of three tabs, see our take on marketing analytics automation.

ChatGPT search compared with Google, Perplexity and Copilot

The same page can do well in one AI engine and be invisible in another, because each one retrieves differently and gives site owners different tools. The table below summarizes what is documented or measured, not guesses about secret weights.

The practical reading: Bing indexing helps two of the four engines directly, crawler access is a separate switch for each one, and only ChatGPT forces you to build your own citation tracking. Perplexity’s higher overlap means classic rankings carry over more cleanly there; with ChatGPT, fan-out coverage matters more than a single position.

What ranking in ChatGPT search costs: a worked example

The technical steps are nearly free. The cost is people’s time for content and outreach, plus an optional tracker. Here is a worked example for a 12-person B2B services firm with a WordPress site of about 60 posts. The hours are our planning assumptions from similar projects, not measured benchmarks, so adjust them to your team.

That totals about 45 hours up front and about 29 hours a month. At an internal cost of $60 an hour, which is an assumption you should replace with your own rate, that is roughly $2,700 to start and $1,740 a month plus the tracker. The arithmetic is simple: 45 x $60 and 29 x $60.

The limits matter as much as the cost. OpenAI documents no way to pay for organic placement, you cannot see impressions, and results shift as OpenAI changes its retrieval. Plan for a quarter before judging results, since content needs to be crawled, indexed in Bing and picked up in fan-out queries before it can be cited. If the budget is tight, do the free technical steps and rewrite three pages, not ten.

Mistakes that keep good sites out of ChatGPT answers

Most sites that never appear in ChatGPT are not losing on content quality. They are losing on one of these avoidable errors.

• Blocking all OpenAI bots in one rule. A blanket block of every AI crawler, often added by a security plugin or a copy-pasted robots.txt, removes you from ChatGPT search. Block GPTBot if you want, but allow OAI-SearchBot.
• A firewall that overrides robots.txt. Cloud bot-protection may serve challenge pages to OAI-SearchBot. Robots.txt says yes, the firewall says no, and the result is no. The curl test in Step 1 catches this.
• Ignoring Bing. Sites with hundreds of pages indexed in Google sometimes have a few dozen in Bing. Check the numbers side by side.
• Using robots.txt to hide a page. If you want a page out of search, use noindex and let the crawler read it, as OpenAI’s FAQ advises.
• Writing for the head keyword only. One page targeting “workflow automation” will rarely be cited for the 30 narrow sub-questions ChatGPT actually searches.
• Clever titles. A witty headline throws away the title-query overlap that AirOps tied to roughly double the citation rate.
• Chasing one platform. Reddit-only or listicle-only strategies are fragile, as the Reddit swings show.
• Measuring with screenshots. Answers vary between runs. Track a fixed prompt set over weeks, not a single impressive screenshot.

A real audit: checking thinkpeak.ai against the list

We ran the first two steps on this site while writing this guide, which shows how quick the technical part is. The live robots.txt at thinkpeak.ai contains only a User-agent: * group that blocks WordPress admin paths, WooCommerce temporary files and plugin folders, plus a sitemap line pointing to Rank Math’s sitemap index. There is no OpenAI-specific rule, so OAI-SearchBot falls under the general group and is allowed to crawl posts.

The curl test with the OAI-SearchBot user agent returned HTTP 200 on the home page, so no firewall rule is challenging it. That clears Step 1. For Step 2, Rank Math is already installed, which means IndexNow is a single toggle rather than a custom integration, according to Bing’s list of supported plugins.

The open items are the slower ones, and they are the same for most small sites: confirming Bing’s indexed page count matches Google’s, adding an AI channel group in GA4, and rewriting older posts whose titles do not name the question they answer. The lesson from our own audit is that the technical gate takes under an hour, while the content and mention work is where the time goes. If you want the content side to scale without losing quality, our guide to content repurposing with AI covers how to turn one researched page into several focused ones.

Other ways into ChatGPT besides organic citations

Organic citations are not the only route. Depending on what you sell, one of these may reach buyers sooner.

Merchant product feeds. E-commerce brands can share a structured product feed with OpenAI. The OpenAI commerce documentation describes a regularly refreshed CSV or JSON feed with identifiers, prices, inventory and media that helps ChatGPT surface products in search and shopping, and says attributes such as rich media and reviews improve ranking and relevance. Instant Checkout requires implementing additional flows and a supported payment provider. If you sell physical products, this is more direct than any content tactic.

Ads. OpenAI’s bot documentation lists OAI-AdsBot, which only visits pages submitted as ads. If paid placement is available in your market, it is the only way to buy visibility; no documented route lets you buy organic citations.

Apps and integrations. SaaS products can build a ChatGPT app so users can act inside ChatGPT. This suits products with a clear action (book, search, create), not service firms.

Classic SEO and Google AI features. Because ranking well in Google still raises the odds of a ChatGPT citation, strong traditional SEO is not an alternative so much as the base layer. Our post on Gemini 3 and SEO covers the Google side.

Choose by business model: product sellers start with feeds, SaaS with apps plus content, and service firms with content and mentions.

ChatGPT search ranking checklist

Use this as a one-page review before and after each quarter of work. Every item maps to a step above.

1. robots.txt allows OAI-SearchBot (and you have made a deliberate choice about GPTBot).
2. A curl request with the OAI-SearchBot user agent returns 200, not 403 or a challenge.
3. Firewall allow rule uses OpenAI’s published searchbot.json IP ranges.
4. Pages you want hidden use noindex, not a robots.txt block.
5. Bing Webmaster Tools set up, sitemap submitted, indexed page count close to Google’s.
6. IndexNow enabled in your SEO plugin or automation.
7. 20 to 50 buyer prompts written down, with their likely fan-out sub-questions.
8. Each money page has a title that names the question, an answer in the first sentence, and a comparison table where relevant.
9. Prices, versions and dates on key pages checked in the last 90 days.
10. Third-party URLs that ChatGPT cites for your prompts listed, with an outreach owner for each.
11. Company description identical across your site, directories and profiles.
12. GA4 AI channel group live and placed above Referral.
13. Server logs checked monthly for OAI-SearchBot visits to key pages.
14. Prompt panel results recorded on a fixed schedule.

FAQ

Can I pay OpenAI to rank higher in ChatGPT search?

Not for organic citations: OpenAI documents no paid route into organic answers. Paid visibility, where available, runs through OpenAI’s separate ads system, which is why OpenAI documents a dedicated OAI-AdsBot that only visits submitted ad pages.

Does blocking GPTBot hurt my ChatGPT search visibility?

No. OpenAI treats them as independent settings: GPTBot is for model training and OAI-SearchBot is for search. You can block training and stay in search by allowing OAI-SearchBot.

How long does it take to show up in ChatGPT search?

Robots.txt changes take about 24 hours to register, per OpenAI. New or rewritten content takes longer because it must be crawled and indexed in the underlying search providers first. Give a content program at least a quarter before judging it.

Does llms.txt help me rank in ChatGPT?

OpenAI’s crawler documentation covers robots.txt, user agents and IP ranges, and does not mention llms.txt. Adding one is harmless, but do it after the steps that OpenAI and the public studies actually support.

Is schema markup a ranking factor for ChatGPT search?

There is no OpenAI statement that schema affects ChatGPT citations. Structured data still helps search engines understand your pages, which supports the Bing and Google indexing ChatGPT relies on, so keep it accurate, but do not expect it to move citations on its own.

Why does ChatGPT cite a competitor with weaker SEO?

Usually because their page answers one of the fan-out sub-questions more directly, with a title that matches it. Run the same prompt, look at which part of their page was used, and check whether you have a page or section that answers that exact question.