An AI content gap analysis finds the questions where AI names competitors instead of you, works out why, and turns that into a list of pages to write.
Short answer: write 20–40 buyer questions, run each one in ChatGPT, Gemini, Perplexity, Claude and Google's AI features, and log who's named and which sources are cited. Then compare the cited sources with your own content. The gap is usually a missing page, a missing use case or a third-party list that leaves you out. Expect a first pass to take an afternoon by hand.
What do you need before you start?
Access to the assistants your buyers use. Usually ChatGPT, Gemini, Perplexity and Claude, plus Google AI Overviews and AI Mode if search traffic matters. Each draws on different sources, so a result in one doesn't tell you about the others.
A neutral setup. Incognito mode doesn't help much here: what personalises AI answers is account memory, custom instructions and chat history. Use a fresh chat, turn memory off or use a clean account, and note your location, because answers can vary by country.
A log. A spreadsheet is fine to start. You'll need a column for each field in step 3.
How do you run the analysis?
Step 1: Write the questions
Cover the decision journey, using the words buyers actually type:
Problem: "Why are my marketing email open rates dropping?"
Solution: "How can I improve email deliverability?"
Comparison: "Mailchimp vs Constant Contact for a small business"
Shortlist: "Best email platform for a 10-person ecommerce team"
Brand: "Is [competitor] good for enterprise teams?"
Comparison and shortlist questions usually matter most, because that's where a recommendation turns into a sale.
Step 2: Run each question in each assistant
Copy the answer word for word or take a screenshot; don't paraphrase. Try one variation of wording ("best tool for…" vs "recommend a tool for…") on your top questions. Run your most important questions two or three times, because answers vary between runs.
Step 3: Log the results
Record one row per question per assistant. A hypothetical example:
Question | Assistant | Date | You | Competitors named | Sources cited |
|---|---|---|---|---|---|
Best email platform for a small ecommerce team | Perplexity | 5 Oct 2026 | Not mentioned | A, B, C | 2 review sites, 1 "best of" list, B's pricing page |
Best email platform for a small ecommerce team | ChatGPT | 5 Oct 2026 | Mentioned 4th | A, B, D | None shown |
A fuller layout is in the AI recommendation gap template.
Step 4: Work out why you're missing
This is the step that makes the analysis useful. For each gap, open the cited sources and ask:
What type of page is it? A review site, a comparison article, a vendor's pricing page, a forum thread, documentation.
What does it cover that you don't? A use case, a price, an integration, a comparison table.
Could you appear on it? Third-party lists and review sites sometimes accept submissions or updates.
In the example above, the fix might be a page for small ecommerce teams with real pricing, plus a request to be added to the "best of" list. The four usual causes are covered in why your brand disappears when AI answers a question.
Step 5: Prioritise and publish
Rank the gaps – the gap scorecard gives a quick 1–5 model – and write the top few pages first. Each one should answer its question in the first two sentences, then add the specifics the cited sources had and you lacked. Publish it, make sure it's indexable, and link to it from related pages.
What makes the analysis more reliable?
Check each assistant separately. Perplexity shows its sources on every answer, Gemini and Google's AI features draw on Google Search, and ChatGPT uses training data plus live search when it browses. Treat these as tendencies to confirm in your own results.
Look beyond competitors. If the answer cites a Reddit thread, an analyst report or a how-to guide, that is the content to match or get mentioned in.
Repeat on a schedule. Re-run the same questions every two to four weeks and keep the dated log. Answers change as models and sources update.
Doing this by hand is fine for 20 questions. Past that, a tracker saves time: dedicated tools such as Otterly.AI and Wellows start at $29 a month, and blogawesome ($49 a month) adds drafting and publishing for the gaps it finds (prices checked 5 Oct 2026).
To skip straight to a first gap list, run the free AI visibility audit. It takes no signup and about 30 seconds, and shows whether AI models recommend you, who they name instead and which content gaps to close first.