Check your ad copy before you launch it
Every ad platform will happily take your money for an ad that was never going to work. The metrics tell you it failed — three days and a few hundred dollars later. Checking the copy beforelaunch costs a minute. Here’s what to check, and how to automate it.
The seven things that decide whether an ad works
A performance marketer reviewing an ad isn’t vibing — they’re checking a short list. These are the seven dimensions Spendict scores on every assessment:
- Hook — does the first line earn the second? Most dead ads die here; a weak hook makes everything after it irrelevant.
- Angle — is there one clear reason to care, aimed at a real pain, or is it a feature list wearing a trench coat?
- Clarity— could a distracted stranger say what’s being sold and for whom after one read?
- Audience resonance— does it speak the audience’s language, or the founder’s?
- Platform fit — a great TikTok script is a bad LinkedIn ad. The same copy scores differently per platform, and it should.
- CTA — is there one obvious next step?
- Compliance— will the platform’s policy review reject it before a human ever sees it?
Check an ad in under a minute
With the CLI:
npm i -g spendict spendict login spendict assess \ --platform meta \ --product "home fitness app, $29/mo" \ --text "Still paying $200/mo for a gym you visit twice?"
You get back scores on all seven dimensions, the predicted failure mode, prioritized fixes, and a deterministic verdict: run (worth spending on), fix_first (a real spine, one specific fixable problem), or kill(structurally broken — faster to regenerate than to repair). The verdict is recomputed server-side from fixed gates, so it can’t be flattered into a pass.
If you’d rather not touch a terminal: connect Spendict to Claude as an MCP connector and just ask it to assess your ad in chat.
Make the check automatic
The manual check is worth doing; the automatic one is worth keeping. If ads come out of an AI workflow — an agent, an n8n flow, a script that drafts variants — put the check inside the pipeline so nothing reaches your ads manager without a run verdict. That’s a single API call per ad, and the first 100 every month are free.
Frequently asked questions
Is this the same as an A/B test?
No — it's what you do before the A/B test. Testing two weak ads against each other still gives you a weak winner. Pre-launch checking makes sure only structurally sound variants earn test budget.
Can it check the image or video too?
Yes — include a creative URL and the assessment covers the visual as well as the copy.
What does it cost to check one ad?
The free tier includes 100 assessments per month with no credit card, so a solo media buyer can usually run entirely on it. Paid plans start at $19/month.
Does 'run' guarantee the ad will perform?
No tool can promise that — performance also depends on offer, audience, and budget. What the verdict guarantees is that the ad has no detectable structural failure: the hook, angle, clarity, CTA, and compliance all clear the gates that kill most ads.