đ Works best with your Meta ad account. Connect it and Jeremy AI reads your live ad text for you. It only reads.
Paste your ad lines and your stated facts. Get one file back: a line-by-line review, safe rewrites for any claim the facts don't support, a clean ad list, a not-provided list and a check table. Nothing is published.
You'll see the SAMPLE ad lines pasted in with the facts and rules, the review built in about 4 seconds on screen (the real wait was about 121 seconds), the file saved with nothing published, the chat summary, the review table with a status, reason and safe rewrite for each line, the clean ad list, the not-provided list and the check table.
What this use case does and who it's for
Ads get written fast. A line slips in that promises too much. "Guaranteed results." "The number one agency." "Our clients double their revenue." If you can't back it up, it's a risk.
Here's the shortcut. Paste the ad lines. Paste the facts you can prove. Jeremy AI checks each line against those facts. If a line matches a fact exactly, it leaves it alone. If it makes a promise, a ranking, a multiplier or a general claim that isn't in your facts, it flags the line and writes a safer version that uses only your facts. Four parts:
A review table. Four columns: line, status, reason and safe rewrite. One row per ad line, in the order you gave them.
A clean ad list. The final version of every line, in order.
A not-provided list. Everything a claim would need that you didn't give, like awards, guarantee terms or disclaimers.
A check table. Three columns: item, source and status. It shows every line was reviewed.
It's for agency owners, media buyers, coaches and anyone who writes ads. It's a first pass, not legal advice. It's drafts only. Jeremy AI doesn't publish anything.
What you need before you start
A Jeremy AI account. Nothing to connect to try the sample.
Your ad lines. The sample is made up: SAMPLE Lumen Growth, a paid-ads agency for gym owners, with six lines (A-01 to A-06).
Your stated facts. Only things you can prove. The sample gives three: a client result (Redwood Fitness Studio went from 6 to 17 new members per month over 90 days, 3 locations, a $4,000 monthly ad budget), the next step (a 20-minute strategy call) and the audience (gym owners with 2 or more locations).
Honest gaps. The sample says rankings or awards, guarantee terms, revenue results, results for other clients, required disclaimers and the ad platform are "not provided."
The status rules. The sample says "ok" means the line matches a stated fact exactly. "Rewrite" means the line makes a promise, ranking, multiplier or general claim that isn't in the facts.
The rules. The sample uses HARD RULES: never add a claim, number, ranking, guarantee, testimonial or detail that isn't in the facts, never calculate a new number, never mention discounts, bonuses, free trials, giveaways, pauses or downgrades, and never address you by name.
How to set it up
You use the default Jeremy AI worker. There's nothing to build.
Open Jeremy AI. From the dashboard, type in the main box. Check the worker selector says Jeremy AI.
Paste the prompt below. Swap your own ad lines and facts in where the sample ones are.
Click Send message.
Wait. The walkthrough speeds this part up. In real time it took about two minutes.
When it says the file is saved, read the chat summary first.
Open the file card and read the review table.
Scroll to the clean ad list, the not-provided list and the check table.
The prompt
This is the exact prompt from the walkthrough. The agency, the client and every number are SAMPLE and made up. Replace them with yours.
Act as my "Ad Claims Check" worker for SAMPLE Lumen Growth, a paid-ads agency for gym owners (all data below is made up). Review the SAMPLE ad lines below against the stated facts and rewrite any line that makes a claim the facts do not support. Create ONE file named ad-claims-check.md using the file tool (reply with a short summary only) with: 1) a REVIEW TABLE with exactly 4 columns (line, status, reason, safe rewrite), one row per ad line in the order given, 2) a CLEAN AD LIST with the final version of every line, in the order given, 3) a NOT PROVIDED list, 4) a CHECK table with exactly 3 columns (item, source, status). STATUS RULES: ok = the line matches a stated fact exactly (copy it unchanged, rewrite cell says "none"); rewrite = the line makes a promise, ranking, multiplier or general claim that is not in the stated facts. A safe rewrite uses only stated facts, copied exactly, and may mention the SAMPLE client by name. HARD RULES: never add a claim, number, ranking, guarantee, testimonial or detail that is not in the facts; the word "guarantee" may appear only in a reason cell that explains why a line was flagged; never calculate a new number; write "not provided" where something is missing; never mention discounts, bonuses, free trials, giveaways, pauses or downgrades; every table has 4 columns or fewer and cells of 12 words or fewer; never address me by name anywhere in the chat reply or the file. Drafts only, nothing is published.
AD LINES (SAMPLE):
- A-01: "Guaranteed 20 new members in 30 days."
- A-02: "Our gym owners double their revenue."
- A-03: "Book a 20-minute strategy call."
- A-04: "Redwood Fitness Studio went from 6 to 17 new members per month."
- A-05: "The number one ad agency for gyms."
- A-06: "Stop wasting money on ads that do not work."
STATED FACTS (SAMPLE):
- Redwood Fitness Studio has 3 locations; new members per month went from 6 to 17 over 90 days; the monthly ad budget was $4,000.
- The next step is a 20-minute strategy call.
- The audience is gym owners with 2 or more locations.
Not provided: rankings or awards, guarantee terms, revenue results, results for other clients, required disclaimers, ad platform.
Prompting tips and how to adapt it
List only facts you can prove. Jeremy AI can only rewrite with what you gave it. If it isn't in the facts, it isn't in the rewrite.
Define "ok" and "rewrite." The prompt says ok means the line matches a stated fact exactly. Everything else gets flagged. No gray area.
Say what a safe rewrite may use. The prompt says stated facts only, copied exactly. That's what keeps new claims out.
Ban new numbers. The prompt says never calculate a new number. No made-up percentages or multipliers.
Control one word. The prompt lets "guarantee" appear only in a reason cell that explains a flag. It can't sneak into a rewrite.
Keep the order. One row per line, in the order you gave them. Easy to check against your doc.
Cap the cells. 12 words or fewer. A short reason is easier to act on.
Adapt it. Swap in any ads: email subject lines, landing page headlines or social posts. Add your own banned words. Keep tables at 4 columns or fewer.
The result to expect
One Markdown file named ad-claims-check.md, titled "Ad Claims Check (SAMPLE Lumen Growth)." With the sample data you should see:
A 4-column review table with six rows.
A-01, "Guaranteed 20 new members in 30 days." Rewrite. Guarantee terms aren't provided and 20 in 30 days isn't in the facts. The safe rewrite uses the Redwood Fitness Studio result.
A-02, "Our gym owners double their revenue." Rewrite. Revenue results aren't provided.
A-03, "Book a 20-minute strategy call." OK. It matches the stated next step. The rewrite cell says "none."
A-04, "Redwood Fitness Studio went from 6 to 17 new members per month." OK. It matches a stated fact.
A-05, "The number one ad agency for gyms." Rewrite. Rankings aren't provided.
A-06, "Stop wasting money on ads that do not work." Rewrite. It implies a results claim the facts don't state.
A clean ad list. The final version of all six lines. The OK lines are unchanged.
A not-provided list. Six items: rankings or awards, guarantee terms, revenue results, results for other clients, required disclaimers and ad platform.
A check table. Each line reviewed, with a source and a status.
The chat reply summarizes which lines were flagged and says nothing was published.
These are sample results from made-up lines. Wording changes from run to run. Read every line before you publish anything.
Check the audience wording
Look at the rewrites for A-02 and A-05. In the sample run, they use the stated audience: gym owners with 2 or more locations. "We run paid ads for gym owners with 2 or more locations" and "Paid ads agency for gym owners with 2 or more locations."
That wording is safe. It comes straight from your facts. But it's narrower than the original lines, which said "gym owners" and "gyms." If you also serve single-location gyms, say so in your facts. If you don't, the narrower wording is correct. Either way, it's your call. Check it before you publish.
Also check two small things in the sample run:
The A-01 rewrite. It says Redwood went from 6 to 17 new members per month. Your facts say that happened over 90 days. A stronger rewrite keeps "over 90 days" so the result has its time frame. Add it if it's missing.
The chat summary count. In the sample run, the reply said "3 lines flagged," then listed four (A-01, A-02, A-05 and A-06). The file is right: four rewrites, two OK. Trust the table, not the count.
Common mistakes and fixes
A rewrite adds a claim you didn't give. Delete it. Remind it: "Safe rewrites use only stated facts, copied exactly."
A line you wanted kept gets flagged. Add the missing fact to your facts list and run it again.
An OK line gets changed. Remind it: "ok means copy the line unchanged and write none."
A new number appears. Delete it. Remind it: "Never calculate a new number."
The word "guarantee" shows up in a rewrite. Delete it. It may only appear in a reason cell.
The audience wording is narrower than you want. Add to your facts who else you serve, then run it again.
You publish a rewrite with a time frame missing. Don't. Check each result has its period.
You publish as is. Don't. This is a first pass. Check your platform's rules and any legal requirements yourself.