Why ChatGPT or Perplexity Miss Your Business or Show Old Facts

ChatGPT and Perplexity may skip your business because they cannot parse it, not because the brand is too small. The problem can happen at several stages: a search crawler cannot reach the page, a firewall blocks the platform's published IPs, the first HTML response lacks the important facts, or other pages on the web contradict your site. Diagnose the cited or retrieved source before changing copy. The two platforms publish different crawler controls, and neither documents a single fixed route from query to citation.

Choose the symptom before changing the site

A missing recommendation and an incorrect address need different investigations. Start with what the answer actually says and retain the cited URLs. This table chooses the next test; it does not diagnose a platform’s ranking model.

How ChatGPT and Perplexity Discover Websites

OpenAI says ChatGPT search may rewrite a user's prompt into targeted queries for third-party search providers. Microsoft is one documented partner, but the current ChatGPT search documentation also names another provider and does not identify one mandatory upstream index. OpenAI separately says OAI-SearchBot is used to surface sites in ChatGPT search. That makes OAI-SearchBot access a confirmed eligibility requirement; a universal Bing-first, ChatGPT-User-second sequence is not confirmed. See the platform-specific ChatGPT optimization guide for the implementation details.

Perplexity documents two agents. PerplexityBot gathers and indexes information so websites can surface in search results. Perplexity-User may visit a page in response to a user's request and include a link in the answer. Perplexity's official crawler documentation recommends allowing PerplexityBot and the published IP ranges, while noting that Perplexity-User generally ignores robots.txt because its request is user initiated. The dedicated Perplexity citation guide covers that platform; this article focuses on diagnosing a cross-platform retrieval or source-conflict problem.

Why AI Systems Skip Certain Businesses

A missing result can come from discovery, access, rendering, or source selection. First confirm whether the relevant crawler requested the page and received useful HTML. Then inspect the exact sources the answer cited. On-page symptoms such as a vague offering, missing proof, or conflicting facts are covered in seven signs ChatGPT cannot find you, including the Business Clarity gap. This diagnosis stays on retrieval and source conflicts so it does not duplicate that checklist.

What Makes a Site Parseable to AI Search

Parseable means the platform can retrieve a useful response and identify the facts needed for the question. Return the company name, offering, location or service area, main answer, and supporting links in the initial HTML where practical. OpenAI and Perplexity do not publish a guarantee that every relevant agent executes client-side JavaScript, so testing the raw response is a safer diagnostic than making a universal rendering claim. Use curl and server logs to inspect what the verified agent actually received.

A successful fetch does not guarantee selection. The page may still be irrelevant to the query, contradicted by stronger sources, too vague to quote, or simply not chosen by the platform's ranking systems. Keep facts consistent between visible copy and structured data, but do not treat JSON-LD or llms.txt as an inclusion switch. Technical Access establishes whether the content is reachable; the separate signs article covers whether the page explains the business well.

Separate Training Crawlers From Search and User Agents

OpenAI's crawler documentation says GPTBot is for content that may train foundation models, OAI-SearchBot is for surfacing sites in ChatGPT search, and ChatGPT-User may visit pages for certain user actions. Blocking GPTBot is a training preference, not a search opt-out. Blocking OAI-SearchBot removes a site from ChatGPT search answers except possible navigational links. OpenAI says robots.txt may not apply to ChatGPT-User, and that user agent does not determine search inclusion. The exact control split is summarized in GPTBot vs ChatGPT-User.

Perplexity makes a similar functional distinction. PerplexityBot is the automatic search crawler and respects its named robots.txt setting. Perplexity-User supports user-requested visits and generally ignores robots.txt. For both platforms, verify the current published IP ranges before allowlisting; a user-agent string alone can be spoofed. After a change, monitor response codes and content returned to genuine requests rather than assuming the rule worked.

Why Does Perplexity Show Outdated Information About My Company?

Perplexity may show outdated company information because the answer cites a stale page, finds conflicting facts across the web, or has not yet reflected a corrected page. The citation is your best starting evidence; the underlying cause remains an inference until you inspect that source and your access logs. Correct the authoritative first-party page, reconcile external profiles that disagree, make the current facts crawlable, rerun the same query, and use Perplexity's official reporting instructions if the stale answer remains.

1. Save the exact query, answer, date, and cited URLs so you can reproduce the problem.
2. Open the cited stale source and identify the exact wrong fact, such as an old address, former product name, closed location, or outdated price.
3. Correct the canonical first-party page first. Keep the visible copy, title, canonical URL, structured data, and modification date consistent.
4. Reconcile conflicting external profiles and pages you control, including business listings, social profiles, partner pages, and old location pages. Ask independent publishers to correct material errors rather than copying the same claim onto more sites.
5. Return the current fact in crawlable HTML, link to it internally, keep the sitemap accurate, and confirm PerplexityBot or Perplexity-User is not receiving a firewall challenge.
6. Rerun the same query after the corrected pages are accessible. If Perplexity still repeats the stale claim, use the flag below the answer or contact support with the query URL, the error, and the expected result.

Run the Diagnosis in the Right Order

Start with evidence from the failed answer, then move outward: cited source, first-party canonical page, conflicting external pages, crawler access, and a repeat query. Do not begin by adding generic FAQs, changing every heading, or publishing a competing article. Those edits cannot correct a third-party page that still carries the old fact. How Answer Patch tests crawler allow-block behavior and live responses is documented in the methodology.

Work through a source correction

Fictional example: a bakery moved, but an answer cites an old directory address. The bakery’s current contact page already lists the new address. Record both URLs, request a correction from the directory owner, and check that old pages on the bakery’s own site point readers to the current location. Do not publish another page repeating the old address just to target the query.

Record type: fictional example
Platform and query: Perplexity; Where is Acme Bakery?
Observed answer date: enter the actual date
Cited source: https://example.com/old-directory-entry
Wrong fact: old street address
Authoritative correction: https://example.com/contact
Action owner: person maintaining the listing
Retest: pending; repeat the same query after source correction
Outcome: record the actual answer and citations

A corrected source, a successful fetch, and a corrected answer are three different results. Mark each separately in the downloadable audit worksheet. The sample report shows how evidence becomes an edit and a verification step; the AI search audit explains the paid deliverable.