How to Spy on Competitors Using the ChatGPT Ads Library

Henry Purchase

Co-Founder

Spying on competitor ads used to be easy. Open the Meta Ad Library, type a brand name, see everything they are running. ChatGPT broke that. There is no public library to type a name into, because ChatGPT ads are served privately inside individual conversations. (If you want the full explanation of why no official library exists, I covered that in the ChatGPT Ads Library 2026 guide.)

This article is not about what the library is. It is about the actual method. A repeatable weekly process you can run to see what your competitors are advertising on ChatGPT, read the patterns, and turn them into ads you ship. I run a version of this every week for our own campaigns and our agency clients. Here it is, step by step.

The method in one line

You cannot search ChatGPT for competitor ads, so you reconstruct the library yourself by simulating the conversations your buyers have, recording which competitors appear, and reading the patterns across what you collect.

That is the whole thing. The rest of this article is how to do each part well.

Step 1: Build your prompt map

You are going to simulate your customers' conversations, so first you need to know what those conversations are.

Open a document and list every high-intent question a potential customer might ask ChatGPT on the way to buying what you sell. Not keywords. Full questions, the way a real person types them. Sort them into three buckets:

Awareness questions. The buyer has a problem but does not know the solution category yet. "Why are my Facebook ads getting more expensive." "How do small agencies find new clients."

Comparison questions. The buyer knows the category and is weighing options. "What is the best CRM for a 10-person agency." "Asana vs Monday for client work." "Alternatives to [competitor]."

Decision questions. The buyer is close to buying and wants a final push or a specific fit. "Best lead-gen agency for SaaS companies." "Cheapest project management tool with client access."

Aim for 50 to 100 questions across the three buckets. This list is your prompt map, and it is the asset the whole method runs on. Build it once, reuse it every week, add to it as you learn.

If you are stuck on what your buyers actually ask, the Focal Hint Generator pulls real questions people are asking ChatGPT in your category, which doubles as a source of prompts for this map.

Step 2: Run the prompts and capture what appears

Now you turn your prompt map into observations.

A few rules that make the data trustworthy:

Use a Free or Go tier account. Ads only show to Free and Go users. If you run these prompts from a Plus or Pro account you will see nothing, because those tiers are ad-free.

Start a fresh chat for each prompt. Prior conversation and prior ad interactions skew what the system serves you. A clean session gives you a cleaner read on what genuinely matches each intent.

Run from the geography you care about. Ads are matched by region. If you sell in the US, observe from the US. A VPN helps if you are testing a market you are not physically in.

Capture the full card every time. Screenshot the sponsored card with the headline, body copy, brand, and the offer. Note the exact prompt that triggered it and the date.

Work through your prompt map on a schedule. Weekly is the right cadence for most categories. Daily if your category is moving fast or you are in an active launch.

Step 3: Log it in a structured way

Screenshots in a folder are not intelligence. You need structure, or you cannot read patterns later.

Set up a simple spreadsheet with one row per ad you observe and these columns:

• Date observed
• Prompt that triggered it
• Intent bucket (awareness, comparison, decision)
• Advertiser (the brand)
• Headline
• Body copy
• Offer / CTA (free trial, discount, demo, guide)
• Creative angle (price, social proof, founder story, FOMO, outcome)
• Landing page it sent to

Every observation goes in as a row. Over a few weeks this becomes the thing you actually wanted: a structured, searchable record of what your competitors are running, when, and against which intents. This is your ChatGPT ads library, the one you built because no public one exists.

Step 4: Read the patterns, not the ads

This is where most people stop too early. They collect a pile of competitor ads, glance at them, and move on. The value is in the patterns across the pile, not in any single ad.

Here is what to look for once you have two or three weeks of rows.

Frequency by advertiser. Sort by brand. Who shows up constantly versus once? A competitor appearing across 30 of your prompts is taking the channel seriously. One appearing twice is dabbling. Frequency tells you who your real competition on this channel is, which is often not the same as your competition elsewhere.

Angle repetition. Sort by creative angle. If four different competitors all open with a price hook, the price hook is converting in your category. If three use the same testimonial format, that format works. Repetition across independent advertisers is the strongest signal you can get, because nobody keeps running an angle that loses money.

Intent coverage. Sort by intent bucket. Where are competitors concentrated, and where is there a gap? If everyone is fighting over decision-stage questions and nobody is showing on awareness questions, that gap is an opening for you to own cheaper, earlier intent.

Offer patterns. Sort by offer. Is the category leading with discounts, free trials, free consultations, or lead magnets? The dominant offer type tells you what this audience responds to, so you are not guessing when you build yours.

Landing page approach. Click through (in a way that does not waste their budget unnecessarily) and note where competitors send traffic. Dedicated landing pages or homepages? Long-form or short? This tells you the conversion bar you need to clear.

Step 5: Turn patterns into ads you ship

Intelligence you do not act on is a hobby. The final step is converting what you found into a launch queue.

For each pattern worth testing, write a one-line brief:

• "Four competitors lead with a price hook on comparison questions. Test a price-led headline against our comparison-stage hint."
• "Nobody is advertising on awareness questions in our category. Build an ad group targeting awareness intent with an educational angle while it is cheap."
• "The category standard offer is a free trial. Test our free guide against a free trial offer to see which our audience prefers."

Notice these briefs reference the pattern, not a specific competitor ad. You are not copying anyone's creative, which is both lazy and lower-converting. You are using the collective behaviour of the market to inform what you test. That is the difference between spying that works and spying that just produces a folder of screenshots.

Then ship. The teams that win ChatGPT ads run two to four new variations a week off this kind of competitor read. The teams that lose collect intelligence they never use.

Make this faster

The honest downside of this method is that it is manual. Building the prompt map is a one-time cost, but running the prompts, capturing the cards, and logging the rows is real weekly work. Budget 4 to 8 hours a week per category if you do it by hand.

That manual labour is exactly what the Focal Ad Library removes. It captures live competitor ChatGPT ads, indexes them by industry and angle, and lets you filter and save patterns without running every prompt yourself, then pushes those patterns straight into a creative brief so the gap between "what is my competitor running" and "here is my variation to test" closes in one place. The method in this article is what Focal automates. Join the waitlist here, or if you would rather have the whole thing run for you, our ChatGPT Ads agency uses this exact process on client campaigns.

A realistic weekly rhythm

Putting it together, here is what a sustainable weekly spying routine looks like once your prompt map exists:

Monday, 30 minutes. Run your prompt map. Capture and log new ads. Most of this is repetition, so it goes fast once you have the habit.

Monday, 15 minutes. Sort the new rows. Note any new advertiser, any new angle, anything that changed since last week.

Tuesday, 15 minutes. Write briefs for any new patterns worth testing. Add them to your launch queue.

Across the week. Ship the variations from the queue into your live campaigns.

An hour a week, and you always know what your competitors are doing on the channel before they realise you are watching.

Frequently asked questions

Can you actually see competitor ads on ChatGPT?

Yes, but not through an official library, because OpenAI does not publish one. You see them by running the kinds of questions your customers ask in a Free or Go tier account and recording which competitors appear in the sponsored cards. This article is the full method for doing that systematically.

Why can't I just search a competitor's name to see their ChatGPT ads?

Because ChatGPT ads are served privately inside individual conversations, not in a public, searchable library like Meta's. There is no name-search function. You have to reconstruct the picture by simulating buyer conversations and observing what appears. The ChatGPT Ads Library guide explains why in full.

How often should I spy on competitor ChatGPT ads?

Weekly for most categories. ChatGPT ad inventory changes constantly, so a monthly check misses too much. If you are in a fast-moving category or running an active launch, daily is worth it. Once your prompt map exists, a weekly run takes about 30 minutes.

What should I record about each competitor ad?

Date, the prompt that triggered it, the intent stage, the advertiser, the headline, the body copy, the offer, the creative angle, and the landing page. Logging these consistently is what lets you read patterns across competitors later, which is where the real value is.

Is it legal to spy on competitor ads this way?

Yes. You are observing publicly served ads as a normal user would see them and recording what you see. That is standard competitive research, the same as browsing the Meta Ad Library. You are not accessing anything private or breaking any terms by looking at ads served to you.

Do I need a tool, or can I do this manually?

You can do it entirely manually with a spreadsheet and a Free tier ChatGPT account. The trade-off is time, roughly 4 to 8 hours a week per category. A tool like the Focal Ad Library automates the capture and indexing so you spend your time on the analysis, not the data entry.

My question is not listed here.

Ask in the FutureProof community or email support@tryfocal.com.