How do I use Gemini to research the Meta Ad Library?
A Gemini workflow for Meta ads research: get the ads with their run times from an index of the Ad Library, then use Gemini for the patterns, the briefs and the rewrites.
Daryl Budiman
Updated · 6 min read
TLDR
Gemini cannot open the Meta Ad Library and takes no custom connectors in the consumer app. Klayson finds the ads; Gemini is excellent at synthesis once they are in front of it.
Get the ads from Klayson (search by meaning, top performers first, free), copy the cards or the agent's answer, paste into Gemini.
Ask Gemini for patterns across Klayson's top performers, then a brief and hooks for your product.
Keep Klayson's dates and version counts in the paste; they are the evidence.
A Gem with the prompts makes Klayson plus Gemini a weekly two-minute routine.
Why Gemini needs the ads pasted in
Gemini writes a good essay about "what works in supplement ads" from memory, undated. The library changes daily and the numbers that matter (start dates, versions) are not in any training set. So the data comes from an index of the library, and Gemini does what it is good at: reading twenty ads and saying what they share.
How do I get the ads data into Gemini?
Gemini's consumer app does not take custom connectors, so it cannot query the index directly. The working pattern is two windows: Klayson for the data, Gemini for the thinking.
- Run the search in Klayson (sign up to see the results). Keep the default Sort: Top.
- Select the cards you want to reason about and copy the text (advertiser, days running, versions, hook, copy). Or ask Klayson's agent the question first and copy its answer, which includes the numbers and the ad ids.
- Paste into Gemini with the prompt below. Gemini is good at the synthesis: patterns across twenty hooks, a brief, a rewrite in your voice.
- For a recurring workflow, save the prompt as a Gem so the instructions persist.
If you use Gemini through the API or an agent framework that supports MCP, the same connector address works there: https://api.klayson.com/mcp.
The workflow
- In Klayson, search the category ("supplement ads") and keep the default Sort: Top. Or ask the agent a question ("which angles perform best for supplements?") and copy its answer, which carries the numbers.
- Copy ten to twenty cards: advertiser, days running, versions, hook, copy, offer.
- Paste into Gemini:
These are supplement ads from the Meta Ad Library with their run time and version count. Group them by angle (offer, problem, proof, story, claim). For each group, tell me the median days running from the data and the pattern in the first line. Do not add ads you remember; use only what is here.
- Then the brief:
Take the three top performers by days and versions. For each, give the structure: hook, angle, offer, proof, CTA. Then write the same structure for a magnesium gummy, with three hooks in the customer's words.
- Save the two prompts as a Gem called "Ad research".
What to keep Gemini honest about
- Ask it to quote the days and versions from the paste. Those are the evidence; a pattern without them is an opinion.
- If it proposes a CTR or a spend, remind it Meta does not publish those for ordinary ads.
- If the paste is small (under 20 ads), ask it to say so in its conclusions.
A Monday routine
Search "what started this week in {category}" in Klayson, copy, paste with "what is new versus last week's paste". Two minutes.