How to Rank in ChatGPT Search: A PR Team Checklist

Updated on Oct 21, 2026

By Ronn Torossian

ChatGPT Search does not rank pages the way Google does. It retrieves candidate pages from the live web, extracts the passages that best answer the question, and cites the sources it used. To get cited, a PR team needs clear, specific paragraphs with the answer first, consistent messaging across owned and earned sources, and coverage in outlets the engines already quote.

PR teams that apply a Google playbook to ChatGPT often wonder why nothing moves. The unit of competition is different: a paragraph, not a page. That changes how a team builds authority, structures content and measures success. The data supports the direction. Muck Rack's May 2026 report found that earned media accounts for 84% of AI citations, so coverage a company earns is what the engines quote most.

This article covers the ChatGPT-specific layer. For schema markup, bot crawlability and structured-answer formatting, read the GEO for PR checklist. For Google's AI answers, read the answer engine optimization guide.

How Does ChatGPT Search Choose Which Sources to Cite?

ChatGPT Search chooses sources in two steps: it finds candidate pages on the live web, then it reads them, extracts the most relevant passages and cites the pages it drew from. The first step resembles a traditional search engine, and the second does not.

In the first step, ChatGPT Search pulls candidate pages by topical relevance, freshness and the authority of the domain. It has been widely reported to draw on Bing's index for this step, so a site's existing search and earned media foundation still matters here.

In the second step, ChatGPT Search does not send the reader to a blue link. It synthesizes an answer and attaches inline links to the pages it used. For a PR team, the content competes for a citation, not a click, and the unit of competition is the paragraph.

Why Does Passage Clarity Matter More Than Page Authority?

Passage clarity matters more than page authority in the citation step because the engine reads the text and quotes the paragraph that answers the question most directly. Google's ranking leans on page-level and domain-level signals such as backlinks, so a mediocre paragraph on a strong domain can still rank. A citation does not work that way.

In my own testing of product-comparison queries, mid-authority sites are often cited next to category leaders, and sometimes instead of them. The common factor is how plainly each cited passage answers the query. Four habits make a passage easy to extract:

• Lead with the answer. The first two sentences of a section should state the claim, with no wind-up.
• Make one claim per paragraph. A paragraph that mixes pricing, features and positioning loses to three paragraphs that each make one point.
• Name entities explicitly. Write "Salesforce" or "HubSpot" instead of "the leading platform."
• Use numbers with a source. "Grew 40% year over year, according to the company's 2025 annual report" can be quoted. "Experienced significant growth" cannot.

How Many Sources Does ChatGPT Search Cite Per Answer?

ChatGPT Search usually cites a handful of sources per answer and lists them side by side, with no first position. That count varies by query, so treat it as an observation from testing, not a fixed rule.

The competitive result is that a brand does not need to beat every rival. It needs to be one of the sources cited. Google gives one brand most of the clicks for a query, while a ChatGPT answer can cite several brands at once. Three consequences follow for PR teams.

• Earned placements in several outlets raise the odds. If Forbes, Adweek and the company blog each hold a strong passage about the brand, the engine may cite more than one of them in a single answer.
• Displacing a competitor is less binary. A brand can share an answer with a rival and still win attention if its cited passage is more specific and more useful.
• Consistent presence across many queries beats dominating a single query.

What Happens When ChatGPT Finds Contradictory Sources?

When ChatGPT Search finds contradictory sources, my testing suggests it favors the clearest and most definite statement, not the most-linked page. OpenAI does not publish its selection rules, so this is an observation from queries, not a documented mechanism.

A well-structured press release with a direct factual claim can win a citation over a longer feature article that hedges the same point. Three practices follow:

• State positions in plain, definite sentences. "5W is an independent PR, creative and GEO agency" can be quoted. "5W is positioning itself as a leader in an emerging space" cannot.
• Keep the same wording across the website, press releases, executive bios and social profiles. A press release, an About page and a CEO's profile that say three different things give the engine nothing consistent to cite.
• Make earned quotes match owned messaging. The closer a published quote is to the website's wording, the easier it is for either source to be cited.

What Is the Difference Between Training Data and Live Retrieval?

Training data is what the model learned before its last update, and live retrieval is what ChatGPT Search pulls from the web at the moment of the question. Each layer answers a different kind of query, so a PR team needs a plan for both.

Questions such as "What is this company?" or "Who is this CEO?" are often answered from training data, which includes a brand's Wikipedia entry, major press coverage and executive bios. A team cannot change that layer on demand, but it can build a consistent, well-sourced presence that later training runs are more likely to reflect.

Questions such as "Best CRM for mid-market software companies" or "Top PR firms for AI companies" tend to trigger a live search, as do time-sensitive questions about launches, earnings, personnel moves and crises. Fresh, well-structured passages matter most for those. A rapid-response statement should be written to be quoted, not only published quickly.

Which Example Prompts Show How ChatGPT Selects Sources?

Three kinds of prompt show how source selection tends to work, and a team can run each one and inspect the citations. The patterns below come from the editor's testing and are not guaranteed outcomes.

For "What is the best PR agency for AI companies?", cited pages tend to make a direct, specific claim about who the agency serves, while pages that say there are many good agencies for technology tend to be skipped. For "How does GEO differ from SEO?", cited pages tend to define both terms in adjacent sentences. For "What happened with a brand's controversy?", cited pages tend to be the most factual accounts, with dates, names and outcomes, so a clear official statement can compete with hostile coverage.

How Does ChatGPT Citation Differ From Google Ranking?

ChatGPT citation differs from Google ranking in what competes, what signals count and how many winners a query has. The table compares the two on the points that matter to a PR team. It reflects the editor's working view, not a published specification from either company.

Does Link Building Still Matter for ChatGPT Search?

Link building still matters for getting a page into the candidate pool, but it appears to matter much less than passage quality once the engine chooses which candidates to quote. A page with thousands of backlinks and a muddled passage can lose a citation to a page with few backlinks and a clear answer.

Three shifts follow for PR teams. Stop treating link counts as a stand-in for AI visibility. Put effort into the writing itself, with tight, specific, entity-rich paragraphs that answer the question in the first sentence. Keep earning placements in credible outlets, because those outlets are the sources the engines already cite, as the Muck Rack data above shows.

What Should a PR Team Check First for ChatGPT Search?

A PR team should run seven checks on its own pages and coverage, starting with the passages a buyer would ask about. Each check can be done without special tools.

1. Audit passage clarity. Read every key page and confirm each section holds one self-contained, factual sentence that answers a buyer's question.
2. Align messaging. Use the same wording for core positioning on the website, press releases, executive bios and earned quotes.
3. Front-load answers. Lead every section and every press release paragraph with the claim, then support it.
4. Build several strong passages. Spread them across owned, earned and partner content so more than one page can be cited.
5. Check the training-data baseline. Ask ChatGPT about the brand with a general knowledge question and note what it already says, then list what is wrong or missing.
6. Test live retrieval. Ask decision-stage questions for the category, record which pages are cited, and note which passages were quoted.
7. Write for extraction as well as engagement. A paragraph that can be lifted cleanly and still reads well to a person does both jobs.

Teams that want an outside check can ask 5W to run a citation audit and build a plan from the results. For the strategic layer above this work, see the AI communications pillar and the generative engine optimization pillar.

Frequently Asked Questions

How does ChatGPT Search rank content?

ChatGPT Search does not rank pages in a list. It finds candidate pages on the live web, extracts the passages that best answer the question and cites the pages it used. The unit of competition is the passage, not the page.

Does domain authority matter for ChatGPT Search?

Domain authority helps a page enter the candidate pool. After that, passage clarity and direct relevance appear to decide whether the page is cited, so a lesser-known site with a clearer answer can be cited alongside a larger one.

How many sources does ChatGPT Search cite per answer?

ChatGPT Search usually cites a handful of sources per answer and shows them side by side. The number varies by question, and OpenAI has not published a fixed count.

What is the difference between ChatGPT training data and live retrieval?

Training data is what the model learned before its last update and tends to shape questions like "What is this company?" Live retrieval happens when ChatGPT Search handles current or decision-stage questions, where fresh and well-structured content has the advantage.

What share of AI citations comes from earned media?

Muck Rack's May 2026 edition of What Is AI Reading? reported that earned media accounts for 84% of AI citations across ChatGPT, Claude and Gemini.

About the author

Ronn Torossian is the founder and chairman of 5W AI Communications, the AI Communications Firm. He is the author of two best-selling editions of For Immediate Release.