Gap Analysis Between Two Documents with ChatGPT Deep Research
Comparing two documents by hand is slow, tedious and easy to get wrong. This Fast Forward episode shows how to use ChatGPT Deep Research to run a structured first-pass gap analysis across contracts, policy drafts, strategy documents or supplier proposals.
What this Fast Forward episode is about
If you have ever spent hours with two documents open side by side, trying to work out what one version has that the other does not, you know the pain. Contract versions, strategy drafts, policy updates and supplier proposals can all hide important differences in plain sight.
This Fast Forward episode shows how to use ChatGPT Deep Research as a structured first-pass comparison tool. The goal is not to remove human judgement. The goal is to get to the important gaps, contradictions and ambiguities much faster.
The value is not just comparing documents faster. The value is seeing what a tired human reviewer might miss.
Why this matters
Document comparison is one of those tasks that feels simple until you are deep inside it. You are not just checking words. You are checking obligations, clauses, assumptions, commitments, contradictions and missing detail.
AI helps because it can hold the two documents together and apply a consistent comparison structure. That makes it useful for a first pass before a human reviewer decides what matters, what is acceptable and what needs escalation.
Two documents can take hours to review properly, especially when sections are similar but not identical.
Missing clauses, weak wording or inconsistent commitments can change the meaning of an agreement or plan.
Deep Research can produce a comparison table that highlights missing provisions, contradictions and ambiguity.
Use AI to surface issues quickly, then apply professional review to decide what is significant.
What ChatGPT Deep Research does here
In this workflow, Deep Research is used as a document comparison assistant. You upload two documents, describe what each one is, then ask for a structured analysis of missing provisions, extra provisions, contradictions and inconsistent wording.
The strongest outputs come when you avoid vague instructions like “compare these documents” and instead tell the system exactly what to look for.
Useful for checking whether a revised version has lost important clauses, commitments or controls.
Useful for spotting new obligations, new wording or added expectations.
Useful when two documents appear aligned but say different things about responsibilities, timing or scope.
Useful when terms, dates, commitments or standards are unclear or inconsistent across the documents.
The five-step workflow
Keep the process simple. The aim is to create a repeatable first-pass review that you can use whenever two documents need to be checked against each other.
Start a new ChatGPT conversation and upload Document A and Document B. Use searchable PDFs, Word documents or plain text where possible.
Select Deep Research or the relevant research mode before sending the prompt so the task is treated as a deeper analysis.
Explain what each document is, its purpose and why the comparison matters.
Request section references, gap type, which document contains the wording and why it matters.
Use the output as the first pass, then check source sections and decide which findings need action.
Scanned image-only PDFs should be converted to searchable text first so the tool can read them properly.
The gap analysis prompt template
The prompt below is the core of the workflow. Fill in the document descriptions before running it.
Best use: treat the result as a first-pass issue register, not a final legal, contractual or policy opinion.
What makes it work well
Prompt structure is everything. Asking for four specific types of difference will usually produce a much better result than asking for a general comparison.
Asking for a table also helps because the output becomes easier to scan, copy into a report, discuss with a stakeholder or use as the basis for a more formal review.
Use Document A and Document B, then explain what each one represents.
Missing content, added content, contradictions and ambiguous wording give the model a clear comparison frame.
References make the output checkable and easier to verify against the original documents.
AI can flag issues, but you still need to decide which findings are material.
The practical payoff
The payoff is speed and focus. Instead of spending an afternoon manually hunting for differences, you can generate a structured first-pass issue list in minutes, then spend your human effort where it matters most.
This is especially useful for consultants, analysts, managers, business owners and anyone dealing with version changes, supplier comparisons, policy updates or client documents.
Compare a proposed version against a previous draft to spot missing or added obligations.
Check whether a new version changes requirements, definitions, responsibilities or controls.
Compare competing proposals against the same requirement set or expected scope.
Find where a later draft has added, softened, removed or contradicted earlier commitments.
What this prepares you for
This Fast Forward episode introduces an important Zero to AI pattern: use AI to create a structured first pass, then apply your judgement to improve the final decision.
That same pattern appears across assistants, panels, dashboards, workflows and the later Learning Labs. The human remains responsible for context, judgement and final interpretation.
Use AI for the first pass.
Let Deep Research surface the gaps, contradictions and inconsistencies. Then use your judgement to decide what really matters.
