How to research competitor ads

Most competitor research produces a folder of screenshots that nobody opens again. It feels like work, it is genuinely interesting, and it changes nothing about what you launch next week, because a screenshot of an ad is not a hypothesis.

The version that works is narrower and duller. You are looking for one thing: which of their ideas have survived long enough to suggest they are working. Everything else is decoration. Do it monthly, keep the output as a written test queue rather than an image folder, and it becomes the cheapest source of creative direction you have.

Six steps to research that produces tests

1. Pick the competitors you actually compete with

   List the brands that show up in your auctions and in your customers' consideration set, not the category leaders you admire. Five to ten is plenty. A brand ten times your size is solving a different problem with a different budget, and their ads will mislead you.

2. Pull their live ads from the Meta Ad Library

   The Meta Ad Library is public and free, and it shows every ad currently running across Meta platforms for any advertiser. Search by brand and pull the full set rather than the first few, because the interesting patterns only appear at volume.

3. Sort by how long each ad has been running

   Run duration is the only performance signal visible from outside. Nobody keeps an unprofitable ad live for four months. Treat the long-running ads as their proven set and treat the ones launched last week as their tests, which are interesting for direction but prove nothing yet.

   This one habit separates useful research from a mood board. The newest, most eye-catching ad in the library is the one most likely to be quietly switched off in ten days. The dull-looking ad that has been running since spring is the one paying their rent. Copy the reasoning behind the survivors, not the aesthetics of the newcomers.

4. Break each surviving ad into its parts

   For every long-running ad, record the offer, the hook in the first three seconds or the first line, the format, the proof used, and where it sends people. You are separating what they are saying from how it looks, because the message is the part that transfers to your brand.

   Landing pages are the half most people skip and the half that explains the ad. The offer in the creative and the offer on the page have to match, and seeing where a competitor sends traffic often reveals a bundle, a guarantee or a price structure that the ad only hints at. If three competitors independently arrived at the same guarantee, that is a market fact worth knowing.

5. Log patterns, not pictures

   Write down what repeats across the surviving set: a recurring objection being handled, a format everyone converged on, a proof type used again and again. A pattern seen across several competitors is a market signal. A single clever ad is an anecdote.

6. Convert the patterns into a ranked test list

   Turn each pattern into a specific test with a hypothesis, written in your own brand's voice and with your own claims. Rank by how cheap the test is against how much it would change if it worked. Research that does not end in a queue of tests was entertainment.

What competitor research cannot tell you

You cannot see their spend, their conversion rate, their margin or their return. For commercial advertisers the Meta Ad Library shows the creative and how long it has been live, not the results. Every conclusion beyond that is inference, and the safe inference is limited to: this ad has survived, so it is probably not losing money.

That limit matters because it caps how much of their strategy you should copy. A competitor running an aggressive discount might be clearing inventory, funding growth from a raise, or operating at a margin you do not have. Copy the structure of what works and price it for your own economics. Copying the creative itself is both an intellectual property problem and a strategic one: you arrive second with their message.

Doing it without the tab management

The manual version means the Ad Library, a spreadsheet, and an hour that never quite gets booked. That is why the research gets done in a burst before a planning meeting and then not again for six months, which is the wrong cadence for a channel where creative fatigue is measured in weeks.

Muze tracks competitors as records rather than browser tabs: list the ones you follow, start a scrape of their live Meta ads, and read back what was found, all from ChatGPT or Claude. Scraping is a write action, so it needs a paid plan, while the free tier covers reading what has already been collected. Setup is the same OAuth flow as everything else on Muze MCP, and the creative ideas that come out of it land next to your own Meta performance in the same conversation.

Prompts for a monthly research pass

• “List the competitors I am tracking and show the ads collected for each.”
• “Start a fresh scrape of this competitor's live Meta ads and tell me when it has finished.”
• “Which of their ads have been running longest, and what do those ads have in common?”
• “Turn the recurring patterns into five creative tests I could run next month.”

Frequently asked questions

Is it legal to look at competitor ads?

Yes. The Meta Ad Library is a public transparency tool published by Meta, and looking at what a company is running publicly is ordinary market research. Copying their creative assets or copy verbatim is a different matter and an intellectual property risk.

Can I see how much a competitor is spending?

Not for regular commercial ads. Spend ranges are published only for political and social issue advertising in the regions where Meta requires that disclosure. For everything else you see the creative and how long it has been running, which is why run duration is the signal to lean on.

How many competitors should I track?

Five to ten that you genuinely compete with. Tracking thirty produces a volume of material nobody reviews. The point is to spot repetition across a set, and repetition shows up quickly in a small, well-chosen list.

How often should I do this?

Monthly is a sensible cadence for most accounts. Creative cycles run in weeks, so quarterly research arrives with its conclusions already stale, while weekly mostly rediscovers the same ads.

What if my competitors barely advertise?

That is useful information on its own, and it means you should widen the set. Look at brands adjacent to your category and at anyone competing for the same moment in a customer's day rather than the same shelf.

Can an AI assistant do competitor research for me?

It can handle the collection and the pattern-finding, which is most of the tedium. Muze can list tracked competitors, scrape their live Meta ads and return what it found inside ChatGPT or Claude. Scraping is a write action and needs a paid plan.

Keep going