If you have ever spent two hours with two documents open side by side, trying to work out what one has that the other does not — contracts, strategy drafts, policy versions, supplier proposals — you know the problem. It is slow, it is tedious, and you still miss things. ChatGPT Deep Research can do it in under 10 minutes.
What ChatGPT Deep Research Actually Is
Deep Research is a feature inside ChatGPT, available on Plus and Pro plans. It was designed to research topics across the web, but it works equally well on documents you upload directly. When you give it two files and a structured prompt, it reads both documents in full, holds them in memory simultaneously, and produces a detailed comparative analysis. It is not skimming — it is reading everything and looking for what you have asked it to find.
The output is structured and specific. It tells you what appears in one document but not the other, where the two documents contradict each other, and where language is inconsistent or ambiguous. It cites back to the source documents so you can check anything it flags.
How to Run the Gap Analysis — Five Steps
Step one: Open ChatGPT and start a new conversation. Upload your first document using the attachment icon, then upload your second. ChatGPT accepts PDF, Word, and plain text files. If you are working from scanned PDFs without text recognition, convert them to searchable PDFs first — Deep Research cannot read image-only files.
Step two: Select Deep Research from the model options before you send your prompt. It appears as a toggle or mode selector depending on your version of ChatGPT. Make sure it is active — standard ChatGPT will give you a much shallower analysis.
Step three: Write your prompt. This is the part most people get wrong. Do not just say ‘compare these documents.’ Be specific. Use the template below.
Step four: Send the prompt and wait. Deep Research takes two to five minutes. It shows you what it is working through as it goes.
Step five: Review the output. You will get a structured analysis — typically in table format if you asked for it — that you can read directly, copy into a report, or use as the basis for a more detailed review. Ask follow-up questions in the same conversation if you need to drill into a finding.
Gap Analysis Prompt Template
I have uploaded two documents. Document A is [describe what it is]. Document B is [describe what it is]. I need you to identify: one, any provisions, clauses, or commitments that appear in Document A but are missing from Document B. Two, any provisions in Document B that are not in Document A. Three, any areas where the two documents contradict each other. Four, any language that is ambiguous or inconsistent between the two documents. Present your findings in a structured table with the section reference, the gap or inconsistency, and which document it appears in.
What Makes It Work Well
Prompt structure is everything. The numbered format — asking Deep Research to look for four specific things — consistently returns more useful output than a general ‘compare these’ instruction. Asking for a table in your prompt produces table output, which is easier to scan and paste into a report.
File quality matters too. Clean digital PDFs and Word documents work best. Scanned images without text recognition will not work.
Expect the output to need some judgement. Deep Research can flag things that are not actually gaps, and it can occasionally miss something subtle. But it gets you 80 percent of the way there in a fraction of the time, and it surfaces issues that a tired human reviewer would likely have skimmed past.
The Practical Payoff
The next time you are facing a document comparison job — two contract versions, two strategy drafts, two policy documents — upload both to ChatGPT Deep Research, paste in the structured prompt above with your document descriptions filled in, and let it do the first pass. What used to take an afternoon now takes the time it takes to make a coffee.
