How we work

Guides, articles and case studies on pantheon.ad are written and signed by Henok Ey. Anani, founder of Pantheon (profile on X). Free tools and data pages are maintained by Team Pantheon. Credibility has to come from what you can check, so every number has a source or comes from Pantheon's data, with the method stated on this page.

• Meta Ad Library. The public library of the ads running on Meta's apps: creative, copy, start date, status, versions.
• EU transparency data. For ads shown in the European Union, Meta publishes reach, age, gender, location and targeting, as required by Article 39 of the Digital Services Act. Reach is an estimate made by Meta.
• Landing pages. The page behind the ad, read when an analysis compares it with the promise of the ad.
• Pantheon's database. The ads Pantheon has collected and checked over time. When we publish a number from it, we state the sample, the period and the date of the extraction.

Nobody pays for months for an ad that doesn't sell, so we start from longevity. The public library only includes pages with at least one ad active for 30 days or more. On each advertiser's card, the "best ad" is chosen with a rule written on the card itself: still active, days running, EU reach, number of versions.

We leave out e-commerce and dropshipping by choice, as well as job ads and political ads. Longevity is a clue, not a guarantee: brand budgets and retargeting keep some ads alive, which is why we cross it with reach, versions and analysis.

Duration goes from the start date shown by Meta to the last time Pantheon checked that the ad was still active. Pantheon rechecks the status of the ads it follows regularly and shows the date of the last check. For an ad that is still running, the real duration is therefore longer than the one we show: our figures are an underestimate, never an overestimate.

• Awareness level. How aware the audience is of the problem, on Eugene Schwartz's 5 levels.
• Persuasion mechanisms. From 0 to 3 per ad, from a closed list of 22, classified by Pantheon's AI. A closed list lets you compare ads, niches and markets.
• Landing page coherence. A score from 0 to 100 on how well the page keeps the promise of the ad.
• EU reach. Taken from Meta's transparency data, not calculated by us.
• Estimated spend. An order of magnitude, not a measurement: estimated impressions multiplied by a CPM benchmark for the niche, adjusted for country and month, with a minimum CPM per country. We show it as a range.

The AI texts are shown as the app returns them. No analysis replaces a test on your audience: it gives you a reason to test one version rather than another.

When a guide quotes Pantheon's data, it states the sample and its limits. Example: the profile of the 849 ads from courses, education and services running for more than 90 days (extraction of 23 October 2026). Those ads entered Pantheon through keyword searches with a minimum-duration filter and through manual curation, so they describe what surviving ads look like, not how many ads survive.

We publish numbers as they come out, with the date. We don't use testimonials, ratings, logos or user numbers we can't show. When an external source doesn't state its method, we don't cite it.

We show other advertisers' ads only to comment on them: reduced screenshots, short quotes and a link to the ad in the Meta Ad Library. We don't host their videos, we don't publish galleries to download and we don't create pages about private individuals. When a face isn't needed for the analysis, we blur it.

If you are the advertiser and want an ad removed from our pages, write to support@pantheon.ad with the link: we remove it.

Every guide shows the date of its last real update; we change the date only when we change the content. If you find an error, write to support@pantheon.ad: we correct it and update the date.