Quote Verification Pass

After AI extracts themes and supporting quotes from qualitative data, run a dedicated verification prompt asking the AI to confirm each quote exists verbatim in the source. This is a separate pass, not part of the initial extraction — AI produces more accurate verifications when explicitly auditing its own work.

The Technique

After the initial analysis, send this follow-up prompt:

QUOTE VERIFICATION

For each quote in the analysis above:
1. Confirm the quote exists verbatim in the source transcript
2. If the quote is a close paraphrase but not exact, flag it and provide the actual wording
3. If the quote cannot be located, mark as NOT FOUND

Output format:
• Quote: [the quote]
• Status: VERIFIED / PARAPHRASE / NOT FOUND
• If paraphrase: Actual wording: [what they said]
• Location: [Participant ID, timestamp, or line number]

What It Catches

Paraphrases — AI combined the gist of what someone said with statistically likely phrasing; the meaning may be close but the words aren’t theirs
Frankenstein quotes — statements sewn together from multiple moments in the interview; looks authentic, isn’t
Fabricated citations — timestamps and participant IDs that look authoritative but reference passages that don’t exist
Confident-sounding errors — AI presents all quotes with equal confidence; verification reveals which are solid

When run on ChatGPT-generated quotes, this frequently reveals that a majority of quotes are paraphrased rather than verbatim, even when the extraction prompt explicitly asked for verbatim quotes.

When to Apply

Before any research readout where quotes will be attributed to specific participants
Whenever the initial extraction asked for short/punchy quotes (high synthesis risk)
For any analysis where specific wording matters (messaging validation, pricing research, emotional resonance)

Also applies to Measure phase: customer feedback analysis, churn interview synthesis, NPS driver analysis.

Notes

This adds ~5 minutes to analysis workflow
Running AI as a self-auditor works better than expecting it to catch errors in the initial pass — verification is a cognitively different task than extraction
If source transcripts are very long, specify which section of the transcript to search in the verification prompt

Quote Selection Rules for AI Analysis — the upstream technique that reduces errors before they happen
Multi-Pass AI Analysis Verification — the broader verification framework that includes this as one component
LLMs Generate, Not Retrieve — root cause mental model

Sources

Caitlin Sullivan — “How to Do AI Analysis You Can Actually Trust” (Lenny’s Newsletter, 2026-02-18)

“A majority of the quotes in the previous ChatGPT output were paraphrased, not original verbatim customer statements. This happened with a request for a small set of quotes, so imagine what happens when you’re digging through 20 interviews.” “Without verification, ‘quotes’ like these end up in your deck attributed to a real participant. Sometimes it’s no big deal. Other times, it’s the difference between product language that strongly resonates and messaging that doesn’t convert.”

Unique contribution: Specific verification prompt template; empirical finding that verification frequently uncovers paraphrases even when verbatim was requested.