AI analysis
Every ad, read for you.
adintel downloads each creative, transcribes the video and has Claude read it: the hook, the angle, the emotions it leans on, who it speaks to and the offer. You skim one panel instead of watching the ad five times.
Start your free trial Read the guide
How it works
What happens to every creative
- Saved onceThe media is downloaded and hashed, so forty ads running the same video become one creative with forty uses.
- TranscribedVideo audio becomes timestamped lines. Music-only videos are recognised and left empty, not filled with noise.
- Read by ClaudeThe model looks at the frame, the copy and the transcript and returns a structured reading, not a paragraph of fluff.
- FilterableEvery tag becomes a Library filter, so you can ask for humor-led UGC with a discount offer and get it in one click.
The AI analysis panel
What the ad is doing, at a glance.
A radar shows the emotional drivers the ad leans on. Under it: the content style, a one-line creative targeting summary, the persona it speaks to, and chips for the angle, production style and offer. A plain-English summary says what happens in the ad and why it works.
- Nine drivers scored 0 to 100: fear, aspiration, curiosity, belonging, urgency, relief, pride, humor and trust
- The six strongest drawn in a fixed order, so the same feeling always sits in the same place
- Content style: testimonial, facts, storytelling, how-to, problem and solution, offer, lifestyle, comparison, humor or announcement
- Labelled as a model's reading of the creative, not measured data
Transcripts
Read a 50-second ad in ten seconds.
Every video with speech gets a timestamped transcript. Click any line and the video jumps there. The first sentence feeds the Hooks tab, so you can compare opening lines across a whole brand without pressing play.
- Timestamped lines, click to seek
- Searchable from the Library and readable by your AI through MCP
- Costs 0.3 credits per video
Winning signal
A score for what the brand keeps paying for.
Each creative gets a winning signal from public facts only: how long it has run, how many ads reuse it, how many copy variants it has, whether it is still live, Meta's own Scaled flag and whether its landing page held. The card shows each part, so you can see why an ad scores high.
- Labelled strong, testing or weak, and sortable in the Library
- Your team's tags, star ratings and notes sit right next to it
- An inference, never presented as ROAS
Library filters
Search the library by idea, not by brand.
Because every creative carries the same tags, the Library becomes searchable by what ads do. Combine filters, save them as presets, and switch to the Hooks view to see the opening lines that match.
- Brand, format, platform, angle, visual format, emotion, offer type, production style
- Runtime, Scaled, AI-made media and the winning signal
- Saved presets per workspace, so the team shares the same views
Good to know
Which model reads the ads?
Claude. Each workspace can run the tagging on the fuller model (1.7 credits per creative) or a lighter, faster one (0.2 credits). Every call logs its tokens and cost.
How accurate is it?
It is a careful reading by a strong model, not a measurement. Treat the radar and persona as a fast first opinion that your team confirms; the panel says the same.
Do my own ads get analysed too?
Yes. Creatives you upload get the same transcript, tags and radar, and stay private to your workspace.
What happens when credits run out?
New tagging pauses behind an add-credits button ($10 for 1,000). Browsing, boards, filters and everything already tagged keep working.
See your competitors' ads read for you
Seven days free, 500 AI credits included.