How to get value from your competitors' Meta ads
Everyone looks at what their competitors advertise. It is public, it is free, and it is the fastest way to learn what a market responds to. This guide covers what to look for, what it is worth to your brand, and why an AI assistant cannot do the looking by itself.
Why it is worth an hour a week
On Meta, the ad is the main thing you still control. The platform picks most of the audience, so the first seconds, the promise and the offer decide who stops scrolling. Most ads lose money, and finding one that pays takes dozens of tests.
Your competitors run those tests in public. Every active ad sits in Meta's Ad Library with its start date and its versions. Reading it tells you four things.
What survives
An ad that has run for months is probably paying for itself. One that vanished after a week probably was not.
What they are testing now
The ads launched this week show the ideas a rival is betting on next.
What they sell, and for how much
The link behind an ad leads to the product, the price and the offer they push hardest.
Where the gap is
An angle nobody in your category uses is a cheap test for you.
Treat these as signals. The library shows no spend and no sales, so a long run suggests a winner and does not confirm one.
How to read one ad
Take one ad apart before you compare fifty. Start with what the library shows for every ad.
Then look at the creative itself. Motion's talk on analyzing competitor ads splits the hook into four parts, a breakdown it credits to the creative strategist Mirella Crespi. Add the angle, the offer, the format and the landing page, and you have eight things to write down.
| Part of the ad | What to write down |
|---|---|
| Hook: the first shot | A face, a product, a problem or a side-by-side comparison |
| Hook: the text on screen | The line that appears in the first seconds |
| Hook: the sound | A voice, music or silence |
| Hook: the feeling | What it makes the viewer feel, such as curious, annoyed or reassured |
| Angle | The one reason to buy that the ad argues |
| Offer | The deal: a discount, a bundle or free shipping |
| Format | An image, a video, a carousel, or a catalog ad filled from a product feed |
| Landing page | Whether the page repeats the ad's promise and its offer |
The last row matters more than it looks. An ad whose landing page repeats its image, its wording and its discount is one a team has worked on. A mismatch often means a test.
From one ad to a weekly habit
The value comes from doing this every week for the same competitors.
- Collect the active ads of five competitors.
- Tag each ad with its hook, angle, offer and format.
- Count. Which angle appears most? Which ads are older than 60 days? Which are new this week?
- Write one brief for your own team, based on a pattern and not on a single ad.
Steps 2 and 3 are where assistants such as Claude and ChatGPT earn their place. They tag two hundred ads in the time a person tags ten. Step 1 is where they get stuck.
Why ChatGPT and Claude cannot do it alone
Ask an assistant what a competitor is advertising today and it will guess, decline, or describe old campaigns. We checked why on 20 November 2026.
Meta turns their crawlers away
Facebook's
robots.txtdisallows GPTBot, ClaudeBot and PerplexityBot from the whole site.A fetch is refused
A fetch tool that asks for an Ad Library page gets a 403 error back.
The library is an app
It loads its results after the page opens, so a tool that only downloads the page finds nothing to read.
Their knowledge is months old
A model learns from data gathered before it was trained, and ads change every day.
An assistant that controls a browser can open the library the way a person does. That works for a handful of ads. It takes minutes, it returns a description in place of data, and it cannot tell you how many ads it skipped.
Where vexqo fits
vexqo does the collecting. Your assistant or your code sends one request and gets the ads back as JSON, read from the public library at that moment. The assistant does the thinking, and vexqo brings it the ads.
Seconds, not minutes
In our test on 20 November 2026, 30 ads for a keyword came back in 3.0 seconds and 100 ads in 14.1. See the speed test.
The whole answer in one call
Up to 100 ads, the library's own total, and a cursor that continues where the answer stopped.
What each ad sells
Ask for it and every ad comes with the product behind its link: the name, price, brand and rating.
Who runs it
The advertiser's confirmed owner, category and follower counts come free with a search.
One request collects a competitor's ads with their products:
GET https://api.vexqo.dev/facebook/ads?page_id=16453004404&limit=100&include_products=true
X-API-Key: YOUR_KEY
Then hand the answer to your assistant with a prompt like this one:
Here are 100 active ads from one competitor as JSON. Tag each ad with its hook, angle, offer and format. Then tell me which angle appears most, which ads have run longest, and one angle they do not use.
A thousand ads cost between $0.19 and $0.25. The prices are on the home page, and the Meta Ad Library API page shows a full record.
Using a coding agent? Connect vexqo to Claude Code, Codex or Cursor and it makes the request for you.
What it will not tell you
No tool sees a competitor's results. Spend, clicks and sales stay private, and vexqo returns null for a number the library does not publish. The exception is an ad delivered in the EU, where Meta publishes its reach.
Use what you learn to choose angles, and make the creative yourself. Copying a rival's ad word for word invites a complaint and teaches your team nothing.
Try it on five competitors
The trial is free and has 500 credits, enough for about 5,000 ads. Collect the ads of five competitors and let your assistant tag them.
Get an API key