Prompt

How do I troubleshoot it when ChatGPT and Perplexity keep citing the wrong pages?

Technology / SEO & AEO Tools1 observationLast seen Nov 5, 2026

Latest observation

Nov 5, 2026OpenAI APIWeb search: off

If ChatGPT and Perplexity are citing the wrong pages, the issue is usually one of these:

  1. They’re not actually reading the exact source you expect

    • They may summarize from an indexed copy, a scraped version, or a different edition of the document.
    • Page numbers can shift between PDF versions, print vs. digital, or different layouts.
  2. The document’s text layer is messy

    • Scanned PDFs, OCR errors, or broken page labels can cause bad citations.
    • If the PDF has no reliable text layer, tools may guess.
  3. The model is “hallucinating” citations

    • Sometimes the answer sounds plausible, but the citation is invented or attached to the wrong claim.
  4. Search and retrieval are pulling the wrong passage

    • Perplexity and similar tools may retrieve a relevant chunk from a nearby page, then cite it as if it directly supports the claim.
    • ChatGPT can do this too if it’s working from uploaded files or browsing results.

How to troubleshoot

1. Verify the source version

  • Make sure you and the tool are using the same exact document
  • Check:
    • title
    • edition
    • publication date
    • PDF filename
    • page count
  • If it’s a web source, confirm the exact URL.

2. Check whether the page numbers are stable

  • Open the PDF and compare:
    • printed page number
    • PDF viewer page number
    • any internal numbering in the document
  • Some PDFs have:
    • cover pages with no visible number
    • Roman numerals for front matter
    • page labels that don’t match the PDF viewer count

3. Inspect the cited passage manually

For each citation:

  • open the cited page
  • look for the exact sentence or claim
  • if the claim is not there, check adjacent pages
  • see whether the citation is supporting a nearby idea instead

4. Ask for quoted evidence, not just citations

Prompt the tool to:

  • quote the exact sentence
  • include a short excerpt
  • identify the page and paragraph
  • explain why that page supports the claim

Example:

“Please answer with direct quotes from the source and include the exact page number for each quote. If the source does not explicitly say it, say so.”

5. Force it to use only one source

If it’s mixing sources, tell it:

  • use only the uploaded PDF
  • do not use web sources
  • cite only this document
  • do not infer beyond the text

6. Reduce ambiguity in your prompt

Instead of:

“What does the report say about emissions?”

Try:

“In the uploaded PDF, where does it explicitly discuss emissions targets for 2030? Give the exact page and quote the relevant sentence.”

7. Check for OCR or scan problems

If it’s a scanned PDF:

  • run OCR again
  • use a cleaner copy
  • convert to searchable text
  • compare the text layer with the visible page

8. Use page anchors or section headings

If page citations are unreliable, ask for:

  • section title
  • heading
  • subsection
  • paragraph quote
  • figure/table number

9. Compare against the document’s own index or TOC

If the answer points to a page that seems wrong:

  • search the table of contents
  • find the section heading manually
  • verify whether the topic belongs there

10. Treat citations as a lead, not proof

Even good systems can miscite. Always confirm the citation in the original document before relying on it.

Best practice prompt

You can use something like this:

“Use only the attached document. For every claim, provide a direct quote and the exact page number. If the quote does not explicitly support the claim, say ‘not found in source.’ Do not infer or use outside sources.”

If you want to diagnose a specific case

Send me:

  • the exact prompt you used
  • the answer you got
  • the source document or a description of it
  • the citation that seems wrong

And I can help you figure out whether the problem is:

  • bad page numbering,
  • OCR/text extraction,
  • retrieval error,
  • or a hallucinated citation.

Observation history

How did Obsurfable measure this prompt?

Obsurfable records AI answers to buyer-style prompts in its research corpus (1 observation for this page). Metrics are distributions over observations, not a single static ranking.

Which AI systems does Obsurfable collect answers from?

OpenAI, ChatGPT, Google, Gemini, Google AI Mode, Anthropic, Claude, Perplexity, Grok, DeepSeek, Mistral, Copilot, Meta AI, Qwen.

How can brands improve their AEO?

Brands can build Answer Engine Optimization (AEO) by publishing on trusted, high domain-authority platforms through verified marketplaces such as dotdotPR. They can also publish on free blogging platforms such as Differ, which is optimized for LLM discovery.

Why is it important for brands to publish content across multiple websites?

Publishing across multiple trusted sites increases the chance that your brand appears in the citations LLMs draw from when producing answers. Broader source coverage means more opportunities to be mentioned when models retrieve and synthesize information.

Want this interpreted for your brand?

Explorer is the free public corpus. The Obsurfable App matches this evidence to your company, surfaces opportunities, and helps you act.