Competitor ad research often produces a folder full of screenshots and very few ideas you can actually test. The useful step is not collecting more ads; it is turning relevant examples into distinct creative hypotheses your brand can support.
Claude can help move that work from research to a first batch of 12 creative briefs. It should not decide what will convert. Treat its output as a structured starting point for your team to review, develop and test.
A practical competitor ad research workflow
Start with a specific question. For example: how are brands in this category addressing a particular buyer problem, and what proof or offer are they using? A clear scope helps keep the research tied to your product rather than drifting into a general collection of interesting ads.
- Define the search. Choose the product or service, market, target buyer and problem you want to investigate. Make the scope narrow enough to guide what you collect.
- Collect relevant ads. Select examples from the Meta Ad Library and give Claude the ads or their links. Keep each source URL beside the notes and analysis so your team can check the original context.
- Clean the set. Remove duplicates and ads for unrelated products. If the examples answer different research questions, split them into separate groups instead of forcing them into one analysis.
- Map the creative choices. Compare openings, promises, proof, visuals and calls to action. Note recurring angles, but do not assume repetition means an angle is winning.
- Look for relevant gaps. Compare the patterns with your product, positioning and customer reviews. Identify needs or objections that competitors address poorly, or claims your brand can substantiate.
- Draft 12 concepts. Ask Claude for a hook, visual direction, proof point and CTA for each concept. Give your designer a brief to develop, not a finished ad to publish without review.
- Plan the tests. Write one hypothesis and identify a control for each concept. Decide how you will assess the idea after it runs, using the campaign objective and relevant business outcomes.
How to get useful output from Claude
The quality of the briefs depends on the quality of the inputs. A pile of unlabeled screenshots gives Claude little context about what is relevant, which details come from the source, or what your brand is allowed to claim.
Provide a short description of your offer, intended buyer, market and known proof points. Include the selected ad examples with their links. Ask Claude to separate observations from interpretations, and to flag missing information rather than fill gaps with invented facts.
A prompt can be as direct as this:
“Review these selected competitor ads for [product category] and [buyer problem]. Keep each source link beside its analysis. First group duplicates and exclude unrelated products. Then compare the hooks, promises, proof, visuals and CTAs. Separate what is directly observable from your interpretation. Using our product details and substantiated proof points below, draft 12 distinct creative briefs. For each, include a hook, visual direction, proof point, CTA and one test hypothesis. Do not invent product claims, customer results or evidence. Flag any brief that needs more information.”
Review the output against the source material. Check that Claude has not mistaken a visual implication for a verified claim, merged details from different ads, or turned a competitor’s promise into something your brand can safely say. Customer reviews can help identify buyer language and recurring concerns, but they do not automatically substantiate a performance claim.
Make the 12 concepts meaningfully different
Twelve briefs are only useful if they represent more than twelve rewrites of the same hook. Ask for concepts that vary in the buyer problem, angle, proof or format, while staying within your offer and evidence. The goal is a set of plausible directions to evaluate, not a target number of ads to publish regardless of quality.
This connects well with a broader creative portfolio approach: different customers may respond to different messages, so one apparent winner should not become your only idea. For more on structuring that range, see Build an Ad Portfolio, Not One Winning Ad. If the team needs a way to assess formats alongside concepts, the creative formats testing audit can help identify where the mix is narrow.
Turn briefs into controlled tests
Before production, write down what each concept is intended to test. A hypothesis should connect a creative choice to a buyer response: for example, whether a specific objection-led opening earns more qualified attention than the current control. Keep the control visible so the comparison has meaning.
Set the measurement plan before launch. Choose the metrics that match the campaign objective, and allow the test enough time and data to support a decision. Avoid declaring a concept successful because it receives early engagement or looks stronger in a small, noisy sample. The aim is to learn which idea deserves further testing, not to confirm a preference.
Creative development still needs human judgement. Claude can organize examples and speed up first drafts, but your team must check the strategy, claims, brand fit and test design. For a related perspective on that division of work, read Why AI Hooks and Human Closers Win Together.
An active ad is a reference, not a result
A competitor ad visible in the Ad Library tells you that the ad is active. It does not tell you whether it is profitable, what it costs to run, or how well it converts. You may not know its audience, objective, landing-page performance or role in the wider campaign.
Use competitor ad research to generate questions and testable directions, not to copy presumed winners. Keep the source links, check every claim, and let your own campaign evidence decide what to develop further.