ChatGPT Ranking Factors: What OpenAI Documents and What the Data Shows

Those studies show patterns, not rules. Each one used a different sample, a different month and often a different ChatGPT model. So the useful question isn't "what are the secret ChatGPT ranking factors?" It's "which claims have proof, and what does that proof cover?"

This post is the short, evidence-first answer. For all seven factors people quote, each graded with a scoring sheet, read our full breakdown of the seven ChatGPT ranking factors. Here you'll find what OpenAI says, what the data shows, and a 10-minute check to run on your own site.

What OpenAI Says About Finding, Ranking and Showing Sources

OpenAI doesn't publish a list of ChatGPT ranking factors. It spreads its guidance across a help article, a crawler page and a publisher FAQ. Put together, they cover three stages.

How ChatGPT finds pages

ChatGPT "may search the web automatically when your question would benefit from current information," according to OpenAI's search help article. It "typically rewrites your query into one or more targeted queries" and can follow up with narrower ones. Those queries go to outside search providers. The article links two providers' privacy policies: Microsoft's and Shopify's.

OpenAI also crawls the web itself. Its crawler page says OAI-SearchBot "is used to surface websites in search results in ChatGPT's search features." Sites that block it "will not be shown in ChatGPT search answers, though can still appear as navigational links." GPTBot, which gathers training data, is a separate choice. Blocking it doesn't remove you from search. Our OAI-SearchBot explainer covers the bot in detail.

How ChatGPT ranks what it finds

Here OpenAI says the least about ChatGPT ranking factors. ChatGPT "ranks search results using multiple factors intended to help users find relevant, reliable information," and "placement is not guaranteed." That's the whole statement for web pages.

Three nearby facts are documented:

1. Personal context shapes the search. With memory on, ChatGPT "may use relevant saved memories when rewriting a search query." It may also use a rough location from your IP address.
2. Ads sit outside the ranking. OpenAI's ads help page says advertisers "have no ability to shape, rank, or alter ChatGPT's responses."
3. Shopping has its own list. On the shopping help page, merchants are "ranked based on factors like availability, price, quality, and whether they are the maker or primary seller." That's the only published ranking list, and it covers products only.

How ChatGPT shows sources

Answers that used search "may include citations," which open the source when clicked. A Sources button shows "cited sources and other relevant links," so the panel can hold more than the pages quoted in the text. The publisher FAQ adds that ChatGPT puts utm_source=chatgpt.com on referral links, so clicks show up in analytics.

For how those pieces turn into an answer, see our walkthrough of how ChatGPT search decides citations.

What the Data Links to ChatGPT Citations

Outside researchers can't see ChatGPT ranking factors directly, so they compare cited pages with pages that weren't cited. Every row below is a correlation, not proof of cause, and each is dated because ChatGPT keeps changing.

Where the studies agree

Three findings line up across publishers. Brands the wider web talks about get named more. Passages that answer early and plainly get quoted more. And recent, maintained pages beat stale ones on average. The August 2026 fan-out data adds a fourth point: ChatGPT now runs many narrow searches, often aimed at one site, so a clear page on your own domain for each buyer fact has more chances to match.

Where they disagree or age badly

Freshness has a wrinkle. Ahrefs' 2025 study found ChatGPT favors newer pages overall, but its 2026 study found that, within a single prompt's results, older and established pages were often the ones cited. Reddit is the clearest case of data going stale: a source that was among the most cited in July 2026 dropped sharply in August.

What nobody has shown

No public study shows that three popular ChatGPT ranking factors, schema markup, HTTPS response speed and a "neutral tone," lift citations for ordinary pages. SE Ranking even found pages with FAQ schema averaged slightly fewer citations, 3.6 against 4.2. The full grades for those claims are in our seven-factor breakdown.

The 10-Minute ChatGPT Source Check

This check covers the ChatGPT ranking factors that are documented or well supported, in the order problems usually block you. Pick one page that should win a buyer question, such as your pricing or comparison page. You need a terminal, a browser and ChatGPT with search.

Our robots.txt tester handles step 1 for any bot. Steps 4 and 6 use the same ChatGPT answer, so run them together.

The fetch command for step 2

This requests the page with OpenAI's documented user agent, saves the HTML, and prints the status code and time to first byte. The second line counts how often a sentence you want quoted appears.

curl -s -o page.html -w "%{http_code} %{time_starttransfer}s\n" \
  -A "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36; compatible; OAI-SearchBot/1.4; +https://openai.com/searchbot" \
  https://yoursite.example/pricing
grep -c "Plans start at" page.html

A 403, a challenge page or a count of 0 is a fail. This only tests rules keyed to the user agent. Firewalls that check IP addresses also need OpenAI's published ranges allowed.

A quick example

Tallyfold is a made-up invoicing and payments app for agencies. Its team runs the check on tallyfold.example/pricing. Step 1 passes. Step 2 returns a 200, but the grep count is 0: the price table loads through JavaScript, so a crawler that doesn't run scripts sees an empty table. That one failure explains more than any content tweak could. The fix is to server-render the prices, then rerun step 2 until the count is above 0.

Step 2 matters most because Vercel's crawler study found that OpenAI's crawlers don't render JavaScript. If the answer isn't in the raw HTML, none of the other ChatGPT ranking factors get a chance.

How to Act on Your Results

Fix failures in the order of the check. Access problems come first, because they zero out everything else. Then make sure the answer lives in the HTML and sits near the top. After that, work on titles and on the third-party pages ChatGPT already cites, which is where the brand-mention correlations point.

Two warnings keep this honest. First, a pass on all six steps doesn't guarantee a citation, since OpenAI says placement isn't guaranteed. Second, one ChatGPT answer is a sample of one. To see whether a fix worked, run a fixed set of buyer prompts several times before and after, as our guide to measuring GEO explains. For the crawler and fan-out details behind this check, our ChatGPT SEO guide goes deeper.

Rankbox fits the content steps, 3 and 4. Answer-Space Research finds the questions buyers ask ChatGPT, and the Citation-Ready Writer writes source-backed articles that open with the answer. Rankbox doesn't track AI citations or change your robots.txt. The Business plan is $49.50 a month with a 7-day trial (pricing), and the free AI search readiness check scans a URL for 12 AI-readiness signals.