Can Claude or ChatGPT analyze competitor ads?

Why does the chatbot need a connector?

Meta's official Ad Library API covers ads about social issues, elections or politics delivered anywhere in the past seven years, and ads of any type delivered to the UK or EU in the past year.¹² A US marketer looking at a US advertiser's commercial ads is outside both. The Ad Library website is open to anyone, but it is a search page built for people.³ This guide uses a connector that reads it for the chatbot. We did not test how Claude's or ChatGPT's built-in browsing handles the Ad Library page.

How this page was made

Drafted by AI from Meta's documentation and from a run of Hopkin's ad library tools on 2026-11-09, which included a read of one Fireflies.ai video ad's transcript and audio description. Hopkin is our product, and the walkthrough uses it. The ads in the example are public Meta Ad Library ads. We quote their copy and link to them, and we report no numbers about them. The read of those ads is ours, not a transcript of a chat.

How do I analyze competitor ads with Claude or ChatGPT?

Five steps. Each one has a prompt you can copy into the chat. Replace the parts in brackets.

1. Connect Hopkin to Claude or ChatGPT

   Connect Hopkin to the chat app you use (setup guide). Searching and reading competitor ads in the Ad Library does not need a Meta ad account connection.⁵⁷ Pricing as published on the pricing page: Starter is $100 a month with 1,500 credits. An ad library search costs 5 credits, tracking a competitor costs 200 credits a month, and reading a tracked competitor's ads is free.⁴

2. Find the advertiser

   Prompt to copy:

   Search the Meta Ad Library for active ads from [brand], country US. For each ad show the primary text, headline, call to action, format and start date.

   Copy prompt

   Hopkin runs this with its meta_ads_search_ad_library tool.⁵ When a brand name matches several Facebook Pages, give the Page ID instead of the name so the search stays on one advertiser.

3. Track the advertiser and pull what is running

   Prompt to copy:

   Track [brand] in Hopkin. Then list its active Meta ads that started since [date], with each ad's start date and link.

   Copy prompt

   Hopkin refreshes the ads of tracked competitors daily, on Meta and LinkedIn.⁶ To answer "how long has this ad been running", ask for the start date and compare it with today's date. A start date shows how long an ad has run. It does not show whether the ad works.

4. Ask for hooks, offers and formats

   Prompt to copy:

   Group these ads by angle. For each angle give the hook, the offer, the call to action, the format and the date it started. Say where an ad's link points when it names a competitor or a comparison page.

   Copy prompt

   The worked example below shows what this step returns on a brand's public ads.

5. Turn the pattern into a brief

   Prompt to copy:

   From these angles, write a one-page creative brief for [our product]: three angles worth testing, the hook for each, and what we would say differently. Mark anything you inferred instead of saw in the ads.

   Copy prompt

   If you also connect your ad account to Hopkin, the same chat can compare your results with the angles you found. Hopkin's tools are read-only: no tool creates, edits or deletes anything in an ad account.

Worked example: Fireflies.ai's Meta ads

We ran steps 2 and 3 on Fireflies.ai's public Meta ads on 2026-11-09 and kept the ads that first appeared since 2026-10-28. Here is our read of what came back. Each row links to the ad's entry in the Meta Ad Library.⁸

What the link tells you

The two price-comparison ads send people to fireflies.ai/talk/wispr-flow-alternative. The ad copy never names a competitor, and the address does. Asking the chatbot to report where each ad points is how you find out who an ad is aimed at.

Video ads: frames, transcript and music

For a video ad, Hopkin's meta_ads_get_competitor_ad tool returns frame images, the audio track, the voiceover transcript and a description of the music, so Claude or ChatGPT can read what the video says and shows as well as its written copy.⁹ Hopkin takes 2 frames a second for videos up to about 60 seconds, spaces them wider for longer videos and lists at most 120 frame URLs per ad. Fireflies.ai's "just say it" ad, started 2026-10-28, is a 23-second vertical video (1080x1920) with a Download button and the primary text "you think faster than you type. so stop typing."¹⁰ In the voiceover, Fireflies opens on price ("Other dictation apps charge for this. Fireflies Talk is unlimited and free on every plan right now.") and then dictates a message as a demo: "Hey, I went through the doc, everything looks good. I left two comments in the last section for you to check." The music is electronic lo-fi hip-hop at a medium tempo, with a notification chime near the end. We read this ad from its transcript and audio description; the frames were available, and we did not use them for this example.

Prompt to copy:

Open this ad's video details. Give me the hook in the first three seconds from the transcript, the offer, the call to action, and what the music and pacing do. Quote the transcript lines you use.

What can't the Ad Library tell you?

For ordinary commercial ads, the library shows no spend, impressions or results.¹ A long run time is a reason to look closer, and it does not prove an ad works. An ad that stopped may have been paused for reasons the library does not show. Hopkin's competitor tracking covers Meta and LinkedIn, not Google or TikTok.