How to validate a product or niche with Meta ads (free method)

Most guides on this topic tell you to buy ads first: set a $10 to $20 daily budget, run a handful of creatives, and read the results. That works, but it costs money before you know whether the market is even there. There's a cheaper first step. Read the ads other people already paid for. This guide shows how to validate a product or niche with Meta ads by reading the public Meta Ad Library, what that evidence can and can't tell you, and how to run the same check in a few minutes from Claude or ChatGPT.

How do you validate a product or niche with Meta ads?

You validate a product or niche with Meta ads by reading what competitors already run in the Meta Ad Library: how many advertisers sell something like it, how long their ads have been live, and whether several brands repeat the same angle. If the answers are "several", "months" and "yes", the niche has paying advertisers, and you can decide what to test yourself. The routine:

1. Write the niche as a search. Use the words a buyer would use ("kids foot care", "personalised pet gifts"), not your internal product name. Pick one country at a time.
2. Search the Ad Library with the country set and the "All ads" category selected.
3. Count the advertisers, not the ads. Ten ads from one brand is one data point; ten brands with one ad each is a market.
4. Check start dates. Note which ads have been live for months and which launched this week.
5. Group the angles. Which pain point, offer or proof does each long-running ad lead with, and do several brands repeat it?
6. Open the landing pages. A product ad that sends people to a quiz, a bundle or a free-shipping page tells you how the offer is sold.
7. Write the verdict in one line: crowded and proven, thin and unproven, or nobody here. Then pick the angle you'd test with your own budget.

Budget an hour for one niche by hand (our estimate, not a measured time). Less if the niche is empty. And an empty niche is an answer too. The mechanics of each filter are covered in our guide to using the Facebook Ad Library; this post is about reading the results as market evidence.

What does a Meta ad actually prove about a product?

A live Meta ad proves that someone is paying to put this offer in front of buyers. It does not prove the offer is profitable. For ordinary commercial ads the library shows the creative, the start date and the platforms, but not spend, reach or results (per Meta's Ad Library help page, checked November 2026). So you're reading behaviour, not outcomes.

Signal in the library	What it suggests	What it can't tell you
Many different advertisers	Buyers exist, and others think so	Whether any of them make money
Ads live for months	Someone keeps paying to run it	Profit, or just a patient budget
Several brands, one angle	The angle resonates with buyers or is copied	Which brand started it
A burst of new launches	A niche heating up	Whether it lasts a season
No ads at all	Little paid demand, or a gap	Which of the two it is

Two cautions, and they matter more than the table.

First, inactive ads aren't shown, so you only see what survived. Here, according to Shopify's Ad Library guide (checked November 2026), US-targeted ads disappear once they stop running. Meanwhile, according to Meta's Ad Library tools page (checked November 2026), EU-delivered ads are archived for one year after their last impression, and social issue, election and political ads for 7 years. The failed tests that would warn you off a niche mostly aren't there.

Second, absence is ambiguous. An empty search may mean an open niche, or that buyers in it don't respond to Meta ads.

How do you search the Ad Library for a niche?

To search the Ad Library for a niche, open it, pick a country, choose "All ads", and type a keyword or an advertiser name, then filter by platform, media type and active status. The same steps work, according to Shopify's guide, for a niche such as vegan condiments (checked November 2026).

Search quality decides whether your evidence is any good. We pulled US Meta ads with Proxy on October 25, 2026, and a loose search for "new development apartments" returned 50 US ads of which only 6 were property ads, while an exact search for "luxury condos" returned 20 property ads out of 30. Loose keywords drag in unrelated advertisers. Three habits help:

• Use the phrase buyers type, and try two or three variants.
• Prefer exact phrases when a generic word has other meanings.
• Discard off-topic ads before you count anything. A pile of noise looks like a big market.

Sorting matters too. Sorting by total impressions skews toward large advertisers, and sorting by relevance can surface ads that all started in the last few days. Run both and say which you used.

How long should an ad run before it counts as a signal?

There's no official number. Meta doesn't publish a threshold, and Shopify's guide only says a long-running ad "may indicate" it has proven effective (checked November 2026). Research on this is thin, and we'd rather say so than invent a rule. We use 60 days as a rough line, which is a rule of thumb and not a Meta figure. An ad that's still live after two months is at least not being switched off in a hurry.

Read run length with care:

• Start date is when the ad began, not proof it ran without a break.
• One brand with a dozen ads is not twelve signals; look at distinct copy and distinct landing pages.
• Big advertisers run everything for a long time, including things that don't work.

For calibration, in our October 25, 2026 pull of the 50 largest-advertiser US results for "med spa", 41 had been running 60 days or more. That's a mature, crowded market where long-running ads are the norm, so run length alone separated little. The full breakdown is in our med spa competitor ads guide. In a niche where only a few ads pass 60 days and most are new, the signal is much stronger.

How do you check the angle and not just the product?

You check the angle by sorting each long-running ad into the pain point it names, the offer it makes and the proof it shows, then looking for repeats across brands. Same product, different angle can be the difference between a flat test and a working one.

For a niche like children's foot care, a useful pass reads each ad and writes down:

• The buyer: a parent, a podiatrist, a grandparent buying a gift.
• The pain: growing out of shoes, flat feet, sizing confusion.
• The offer: a bundle, a free fitting guide, a subscription.
• The proof: reviews, a clinician, before-and-after photos.
• The format: video, carousel (the library tags these DCO), single image.

Then ask which pain appears in three or more brands. That's your test angle. A pain nobody mentions may be a gap, or may be a pain buyers don't have; only test it if you have a reason to think it's real. For turning those notes into a brief, see writing a creative brief from competitor ads.

How do you do this faster with Proxy?

Proxy is a free MCP server that gives Claude, ChatGPT and Cursor live access to the Meta Ad Library (and Google, LinkedIn and TikTok in the EU), so you can ask for the whole check in one message. Product and niche validation is a common reason people connect it: in the 30 days to November 2, 2026, at least 40 of the 163 Proxy users who stated a goal (about 25%) were researching product or niche validation (agent-stated goals, keyword matched, so a lower bound).

Try asks like these:

Search Meta ads for "kids foot care" in the US, active only. List the advertisers, how many active ads each has, how long the oldest has run, and the format. Skip anything that isn't about children's feet.

Of those, take the five longest-running ads. For each, give the pain point, offer and proof, and say which pain appears in three or more brands.

I want to sell this in the UK. Run the same search for GB and tell me which advertisers are missing from the UK list.

Proxy searches by keyword across advertisers, returns ranked candidates when names collide, and can transcribe a video ad so the spoken offer is on the page. It's free to start: 20 credits a day, no credit card, and an ad-library call costs 1 credit. Pulls are capped, not the whole library, and there is still no spend or results data, because no public library shows it for commercial ads. Set it up from the ChatGPT and Claude setup guide.

What should you do after the ads say yes?

Test cheaply with your own money, because a library only shows what others did. The usual next step is a landing page or pre-sale page and a small budget against the one or two angles you found, judged on clicks, sign-ups or orders. The ad evidence narrows the angles you test and spares you the ones every brand has already worn out. It doesn't replace the test. I'd treat it as a filter on your ideas, never a green light.

Check what the library can't show before you commit: margin, shipping and returns, supplier reliability, and whether the platform's rules allow the product. If you also need the wider method, the pillar guide to spying on competitors' ads covers every library, and the competitor ad audit template gives you a sheet to record what you find.