How Does ChatGPT Choose Which Sites to Cite?

ChatGPT chooses which sites to cite by rewriting your prompt into about two search queries, pulling pages from its search providers, and citing roughly half of what it retrieves. Ahrefs found the page whose title matches the rewritten query gets picked most. For best results, write titles and headings as the question.

Key Points

StageWhat ChatGPT doesWhat you control
Decide to searchRuns a live search on 34.5% of promptsBrand mentions for the other 65.5%
Rewrite the promptSends about 2 shorter queries to its providersTitles that match those queries
Retrieve pagesPulls results from third-party search indexesCrawler access and indexing
Pick what to citeCites about 50% of retrieved pagesWhere the answer sits on the page

How ChatGPT Search Works in Four Steps

ChatGPT search works in four steps: it decides whether to search, rewrites your prompt into search queries, retrieves pages from its search providers, and cites the ones it used. OpenAI describes the middle steps in its own help center, and independent studies have measured the rest.

  1. Decide. ChatGPT searches the web when it judges the question needs current information. Otherwise it answers from what the model already learned.
  2. Rewrite. OpenAI’s help page on searching the web with ChatGPT says it “typically rewrites your query into one or more targeted queries” and sends those to its search providers.
  3. Retrieve. The providers return results, and ChatGPT opens a set of those pages to read.
  4. Cite. ChatGPT writes the answer and links the pages it drew from. The same help page says it ranks results “using multiple factors intended to help users find relevant, reliable information” and that placement is not guaranteed.

That last line is all OpenAI says about ranking. There is no published citation algorithm, so everything past step two comes from people measuring ChatGPT’s behavior from the outside, and it is what my steps to rank in ChatGPT are built on.

So read the numbers in this guide as patterns from large samples, not a spec sheet. They are still enough to act on, since the same few patterns show up in every study I have read this year.

Does ChatGPT Search the Web Every Time?

No, ChatGPT does not search the web every time. Semrush’s clickstream data shows it turned on web search for 34.5% of prompts as of February 2026, down from 46% in late 2024. The other two thirds of answers come from the model’s memory, with no citations at all.

That figure comes from Semrush’s analysis of 17 months of ChatGPT traffic, built on more than a billion lines of US clickstream data. It is a vendor panel, not OpenAI’s own count, but it is the best public number there is.

For a store, that split means you are playing two games. When ChatGPT searches, your pages compete to be retrieved and cited. When it does not, the answer depends on what the model absorbed about your brand during training, which is why some brands get named in answers that carry no links at all.

Shopping questions lean toward the search side. Prices, stock, and “best X in 2026” are exactly the kind of current information that triggers a lookup, so product and buying-guide pages get more chances at a citation than a brand story page does.

Citations only exist in the third of answers where ChatGPT searches, so you need pages for the search and mentions for the memory.

Where Do the Search Results Come From?

ChatGPT’s search results come from third-party search providers plus OpenAI’s own crawler, OAI-SearchBot. OpenAI’s help center links Microsoft and Shopify as providers in its privacy disclosures, which makes Bing’s index the one you can do something about today.

Does ChatGPT still use Bing?

ChatGPT still uses Microsoft as a search provider, so Bing matters. Most store owners have never opened Bing Webmaster Tools, and submitting a sitemap there takes an afternoon. I cover that setup in does ChatGPT use Bing.

But Bing is not the whole supply. OpenAI says it uses multiple providers and does not list all of them, and its own crawler adds pages on top. Treat Bing as the door you can see, not the only door.

Do you need to rank on Google or Bing to get cited?

You do not need a top ranking on Google or Bing to get cited by ChatGPT. Ahrefs tested 15,000 long-tail queries and found only 8.0% of ChatGPT’s in-text citations ranked in Google’s top 10 for the same query, and 8.1% ranked in Bing’s top 10.

That Ahrefs study of AI citation overlap also found that, across AI assistants, 80% of cited pages did not rank anywhere in Google for the original prompt.

The reason is the rewrite step. ChatGPT is not searching the words your shopper typed. It is searching its own shorter queries, and the pages that rank for those queries are the ones it reads.

So an old piece of advice needs updating, and it is the same reason you don’t need to rank in Google to show up in ChatGPT. A year ago the common line was “rank in Bing’s top 20 and you are in.” Today the better target is ranking for the narrow queries ChatGPT writes, which are often questions with almost no search volume of their own.

Which Retrieved Pages Get Cited?

About half of retrieved pages get cited: Ahrefs studied 1.4 million ChatGPT prompts and found 49.98% of the URLs ChatGPT retrieved ended up as citations. The strongest signal was how closely the page title matched the query ChatGPT wrote for itself.

What Ahrefs measuredResult
Retrieved URLs that got citedAbout 50%
Regular search results cited once retrieved88.46%
Reddit pages cited once retrieved1.93%
Pages with readable, natural-language URLs cited89.78%
Pages with opaque URLs cited81.11%
Average age of a cited page500 days

Those figures are all from the Ahrefs study of why ChatGPT cites one page over another, and the last row ties into content freshness and AI citations. Other vendors measuring different samples have put the cited share much lower, around 15%, so treat 50% as the top of the range.

Whichever number is right, the takeaway holds. Getting retrieved is the hard part, and a page that gets retrieved has a fair shot at the citation.

Also look at the title result. If ChatGPT rewrites “what’s a good carry-on for a week in Europe” into “best carry-on luggage one week trip,” the page titled close to that second phrase has the edge. That is why I write titles and H2s as the short question a model would search, not as clever headlines.

The age row matters too. An average cited page is 500 days old, so you do not need to publish daily, but pages left alone for years do slip.

Put the Answer in the First Third of the Page

Putting the answer in the first third of the page gets it cited more often. Kevin Indig’s analysis of 18,012 verified ChatGPT citations found 44.2% came from the first 30% of the text, 31.1% from the middle, and 24.7% from the final third.

His Growth Memo study of how AI pays attention also found cited passages were almost twice as likely to use definitive language, 36.2% against 20.2%. A sentence that commits gets quoted, and a sentence that hedges gets skipped.

That matches how retrieval works. ChatGPT is not grading your whole page. It is looking for a passage that answers one of its queries, and it finds that passage faster when it sits under a clear heading near the top.

So build each page like an outline:

  • One question per heading, phrased the way a shopper would ask it.
  • The direct answer in the first sentence under that heading, with the number or name in it.
  • The proof right after, then the details.
  • No warmup paragraph before the answer.

This restructuring is the first thing I do on a client’s key pages in my AI search services for Shopify stores, before any new content gets written. It costs nothing but editing time.

Write every section so it still makes sense if it is the only part of the page ChatGPT reads.

Let OpenAI’s Crawlers In

Letting OpenAI’s crawlers in is the one hard requirement for a citation. OpenAI’s crawler documentation says sites opted out of OAI-SearchBot “will not be shown in ChatGPT search answers,” and a single robots.txt line is enough to opt out by accident.

BotWhat it doesWhat blocking it costs you
OAI-SearchBotSurfaces sites in ChatGPT searchYou disappear from ChatGPT search answers
GPTBotCollects content for model trainingYour pages stay out of future training data
ChatGPT-UserFetches a page when a user asks for itA user’s direct request may fail

The OpenAI crawler documentation lists each bot with its own robots.txt name, so you can allow search and still say no to training. It also notes a change takes about 24 hours to reach search results.

On Shopify, check two places before you decide whether to block AI crawlers at all. Open yourstore.com/robots.txt and look for any Disallow rule under those bot names. Then check whether an app or a firewall setting is blocking AI bots at the network level, since that kind of block never shows in robots.txt.

Once you have looked, decide on purpose. For a store that wants ChatGPT traffic, the answer for OAI-SearchBot is always allow. Training is a separate choice, and saying no to GPTBot does not remove you from search answers.

How ChatGPT Picks Products for Shopping Answers

ChatGPT picks products for shopping answers from structured product data, not from ads. OpenAI’s shopping help page says product results “are selected independently by ChatGPT and are not ads,” and that merchants are ranked on availability, price, quality, and whether they are the maker or primary seller.

That OpenAI page on shopping with ChatGPT search names the inputs too. They include structured metadata such as price and product description from first-party and third-party providers, plus what the model already said before it looked at any search results.

For Shopify stores there is a head start. The same page says Shopify merchants’ product data is already integrated into ChatGPT through Shopify Catalog. So your titles, descriptions, prices, and stock status in the Shopify admin are the feed, and they shape how ChatGPT picks products to recommend.

That makes messy product data a ranking problem. A vague title, a missing material, or a variant that shows in stock when it is not all go straight into what ChatGPT has to work with. The setup steps are in my guide to Shopify and ChatGPT shopping.

Being the maker helps as well. If you sell your own brand, ChatGPT’s stated merchant ranking favors you over a reseller listing the same item, which is one of the few places a small brand gets a built-in edge.

Check Who ChatGPT Cites for Your Category

Checking who ChatGPT cites for your category takes about an hour with a spreadsheet and ten prompts. You are looking for the pages that keep showing up, since those are the ones your pages have to replace.

Start by writing ten questions a shopper would ask before buying what you sell. Mix them up: two “best” questions, three comparisons, three how-to questions, and two about a problem your product solves.

Then run each one in ChatGPT with search turned on, and log every cited page. Note the site, the page title, and whether it is a store, a publisher, a forum, or a brand’s own help page.

After ten prompts a pattern shows. Usually five or six sites carry most of the citations, and at least one of them is a page you could beat with a clearer answer. Compare its title to the question you asked, and you will often see the title-match effect from the Ahrefs study with your own eyes.

Run the same ten prompts again next month. The list will have changed, and the sites that appear both times are your real competition. For your own traffic, ChatGPT tags its referral links, so the visits it sends show up in GA4, and my guide to tracking AI traffic in GA4 covers the report.

One caution on reading the results: a single run is a sample. Ask the same question twice and the sources can differ, so look for repeats across runs before you change a page.

Can a Small Store Get Cited by ChatGPT?

Yes, a small store can get cited by ChatGPT. Semrush counted about 170,000 unique domains receiving ChatGPT referral traffic in February 2026, up from roughly 71,000 in October 2024. The count peaked near 260,000 in October 2025, so the pool moves, but it is far wider than a Google top 10.

The opening is in the rewritten queries. A prompt like “running belt that fits a large phone and doesn’t bounce” becomes a narrow search with maybe five pages that answer it well. A focused store can be one of the five.

That is the pattern I see with clients. The wins come from specific questions where the store knows more than a general retailer does, and my playbook for small brands in AI answers starts with picking those questions.

One concession before you start: AI citations are not stable. ChatGPT’s source mix changes every few months, and a citation you earn in the spring can be gone by fall. Coverage across many narrow questions is what keeps you in the answers when one page drops out.

Want ChatGPT Sending Shoppers to Your Store?

  • Check robots.txt and your firewall for anything blocking OAI-SearchBot.
  • Rewrite the titles and H2s on your top ten pages as the short questions a model would search.
  • Move each answer to the first sentence under its heading.
  • Clean up product titles, descriptions, and stock status in Shopify, since that data feeds ChatGPT shopping.
  • If you want it done for you, book a call and I will run live prompts for your category, show you who ChatGPT cites today, and name the three fixes I would start with. My AI search services pick up from there.

Sources

Similar Posts