ChatGPT Deep Research for Competitive Intelligence

#01How does deep research work?

You describe the question. ChatGPT proposes a research plan, which you can edit: scope, sources, constraints. It then searches, reads and assembles a report with citations, and you can interrupt and redirect while it works. When it finishes you get a notification and can share or download the report as PDF or DOCX.

Sources can include the public web, uploaded files and connected apps, which deep research uses only for read actions. Under Sites → Manage sites you can restrict the run to chosen sites or tell it to prioritise them.

#02What competitive questions suit it?

#03Which prompts work?

COMPETITOR TEARDOWN
Research Competitor A ([domain]). Cover: target audience,
positioning statement (quote their homepage), pricing model
and published prices, key launches in the last 12 months
with dates, top 3 praised and criticised features from
review sites. Prefer primary sources: their site, docs,
changelog, press releases. Mark anything older than 12
months. Return a table plus a one-page summary.

CATEGORY SCAN
Map the [category] vendors that sell to [audience] in
[region]. For each: positioning in one line, pricing model,
notable integrations. Only include vendors with a live
website. Cite each row.

MESSAGE GAP
Using the teardown above and our positioning [paste], list
claims we make that no competitor makes, claims everyone
makes, and claims competitors make that we can't match.

#04How do you verify the report?

• Open every source behind a number or a date.
• Check the page date; competitor sites change.
• Prefer the competitor's own pages over third-party summaries.
• Delete any claim you can't trace. Don't "keep it for now".

Our view: deep research gets you to a solid first draft of a competitor file in one sitting. The hour you spend verifying it is what makes it usable.

#05Worked example: Ledgerline's competitor file

Ledgerline is a fictional accounts-payable tool. Details are illustrative.

Ledgerline ran the teardown on Competitors A, B and C with their domains prioritised. The report gave a clean table of positioning and pricing models. Verification found one problem: a "new" feature for Competitor B came from an 18-month-old press release. The message-gap prompt showed every competitor claimed "save time", while none mentioned audit trails, which Ledgerline's customers valued. That became the next campaign angle.

#06What should you check before the next lesson?

• Run one teardown with priority sites set.
• Verify every number in the report against its source.
• Store the verified file in your brand project and date it.
• The Claude track covers research in Claude competitor and SERP research.

Next: ChatGPT Work: marketing tasks it can run end to end. Or return to the ChatGPT for Marketers track.

#07Common questions

How long does a deep research run take? It varies with the question. Runs take longer than a normal chat, and ChatGPT notifies you when the report is ready.

Can deep research use my internal files? It can use files you upload and read-only access to connected apps you allow, alongside the web. Check what your workspace permits.

How many deep research runs do I get? OpenAI does not publish fixed numbers in the help article. Allowances vary by plan, and the monthly allowance resets 30 days after first use.