Free ad creative checker
The ad creative checker scores one ad before you spend on it. Paste the primary text, headline and button, pick static or video, and get a scorecard in seconds: hook, offer, call to action, audience, proof, urgency and policy risks. A decision model answers each check, and code counts the facts. It is free and needs no signup.
Version 1 reads text only. A pre-flight signal, not a prediction. Last updated 15 November 2026.
Check your ad
Version 1 checks the text of one ad. Images and video files are not uploaded.
Want a human to read your market?
A senior growth practitioner reads your market and your ad by hand, with the same system that ran on 54 live ads, and hands you decisions and briefs.
The base: $1,200, delivered in 5 business days
- Capture of 60 live competitor ads in your category and country from public ad libraries, static and video.
- Every ad tagged with a taxonomy drafted for your market, and a market snapshot per track: what the category runs, what is crowded, where nobody stands.
- A written verdict by Vlad Yasinski: five decisions with the numbers behind each.
- Six production briefs: three for the static designer, three for the motion team with three hooks each.
- Your own ad from the checker, read by hand against the market.
Limits of the work
- An ad library shows what runs, never spend or results.
- The research finds gaps and writes briefs. The live test decides.
The same work is published on a real market: see the run on 54 live ads. Prices in US dollars, excluding taxes.
Calculator
What you get
- Your category
- 60 live ads
- Tagged market
- Human verdict
- 6 briefs
60
live ads read and tagged
5
decisions with the numbers behind each
6
production briefs
Day 1
Capture
Day 2
Tagging and snapshot
Day 3
Human read
Day 4
Briefs
Day 5
Delivery
What the ad checker looks at
The ROIMOBI ad creative checker is a free tool that scores one ad before launch. You paste the primary text, headline and button, and the checker returns a scorecard for hook, offer, call to action, audience, benefit, social proof, urgency and policy risks. It reads text only in version 1, needs no signup, and treats the result as a pre-flight signal.
| Check | A pass means | Why it matters |
|---|---|---|
| Hook specificity | The opening line names the audience and a concrete promise | The opening line decides whether the rest gets read, and a line that fits any product gives no reason to stop. |
| Offer clarity | The ad states what the viewer gets | A viewer who cannot tell what they get has no reason to tap. |
| Price or free terms | A price, a free trial, a discount or free use is stated | Terms stated up front let the viewer judge the offer before the tap, so the tap carries intent. |
| Call to action | The ad asks for one specific next step | One named next step gives the viewer a single thing to do, and mixed asks split attention. |
| Audience named | The ad says who it is for | Naming the reader lets the right people opt in and the rest scroll on, so clicks come from better-fit viewers. |
| Concrete benefit | The ad names a concrete result or use | A result the viewer can picture does the work that a feature list leaves to them. |
| One main idea | The ad argues one main point | One point is easier to read at scroll speed than four. |
| Product in the first 3 seconds | The first three seconds show or name the product | Viewers who leave early still know what was advertised. |
| Social proof | Reported as present or absent: ratings, user numbers, awards or other people's results | Proof from other people can lower the risk a viewer feels, as long as you can back it up. |
| Urgency | Reported as present or absent: a deadline, a countdown or limited supply | A real deadline can move people who already want the offer, and a false one costs trust. |
| Opening type | Reported only: the kind of opening, shown when the model is 70% sure or more | Lets you compare variants by the kind of hook they use |
| Policy flag: Guaranteed results | Clear: No guaranteed outcome found | A promised outcome is a claim a reviewer can ask you to support. |
| Policy flag: Health or body claim | Clear: No health or body claim found | Claims about the body or health often get strict review on ad platforms. |
| Policy flag: Income or savings claim | Clear: No income or savings claim found | A specific amount someone will earn or save is a claim a reviewer can ask you to support. |
| Policy flag: Addresses a personal attribute | Clear: No direct reference to a personal condition of the viewer found | Copy that tells the viewer they have a condition can read as intrusive, and platforms review this style. |
| Policy flag: Before and after claim | Clear: No before and after framing found | Before and after framing is often restricted in health and body categories. |
| Policy flag: Best or number one claim | Clear: No best or number one claim found | A best or number one claim needs a source you can show if a reviewer asks. |
| Counted facts | Length against the placement guidance, words in capitals, exclamation marks, emoji, links, price patterns | Code counts these, so they are exact, not judged |
A check shows Pass, Fix or Review. Review asks you to read the line yourself. Signals show Present or Absent and are not graded. A raised policy flag links to the platform's own page, and the platform decides: for Meta, Prohibited Commercial Practices, Health and Wellness, Privacy Violations and Personal Attributes and Meta Advertising Standards; for other platforms, the TikTok Advertising Policies and Google Ads policies. Links checked on 15 November 2026.
Length guidance comes from Meta's ads guide for Facebook Feed, which lists 50 to 150 characters of primary text and a 27 character headline as text recommendations (read on 15 November 2026). Other placements and platforms differ, so the checker shows your counts next to that guidance and treats it as a reference, not a limit.
How the check works
1.You paste the ad
Choose static or video, then add the primary text, headline, button label, an optional offer line and, for video, the hook of the first three seconds as text.
2.Jev answers narrow questions
Each check is a yes or no, a pick-one or a score question. Jev, a decision model by TypeSafe AI, returns probabilities and writes no text. Code counts length, capitals, exclamation marks, emoji, links and price patterns.
3.Code decides and shows fixes
Fixed thresholds turn each probability into a status: yes at 70% or more, no at 35% or less, and anything between goes to review. A fixed library supplies the fix for each status, so the same failure shows the same suggestion.
Each check is a narrow question that Jev, a decision model by TypeSafe AI, answers with a probability. Code then turns the probability into pass, fix or review using fixed thresholds, and shows a fixed fix suggestion for each failed check. Because code applies the thresholds and counts length, capitals and exclamation marks, the same ad text produces the same counted facts on each run. The thresholds are the vendor's starting values and are not calibrated on labelled ads.
The score from 0 to 100 adds three parts. Clarity gives 60 points: offer 15, call to action 15, benefit 10, audience 10 and hook specificity 10, with half points for a check in review. Hygiene gives 15 points for lengths inside the guidance and a low count of capitals and exclamation marks. Policy gives 25 points, minus 8 for each raised flag and minus 3 for each flag in review. It is a checklist score, not a forecast.
A worked example
Sample ad written for this page. Not a real advertiser or client. The scorecard below is the real output of the checker for this text: Jev (jev-1.14.0) answered on 15 November 2026, and code applied the thresholds and the fix library.
- Type
- Static
- Primary text
- Speak Spanish on your next trip, guaranteed! Ten minute lessons built around the phrases travellers use most, with a voice coach that checks your accent. Start today.
- Headline
- Learn Spanish fast
- Button label
- Learn more
- Offer line
- First week free, then $9.99 a month
Score 68 of 100 (checklist score, not a forecast). 4 pass, 0 to fix, 3 to review.
| Check | Result | Probability | Fix shown |
|---|---|---|---|
| Hook specificity | Review | 64% (level 1.7 of 3) | The opener covers part of the picture. Add the missing half, the audience or the promise, in the same line. |
| Offer clarity | Pass | 94% | None |
| Price or free terms | Pass | 99% | None |
| Call to action | Review Borderline | 69% | The ask is there but loose. Replace generic wording such as "learn more" with the step itself. |
| Audience named | Review | 41% | The audience is implied. Say it in words, by role or situation. |
| Concrete benefit | Pass | 94% | None |
| One main idea | Pass | 94% | None |
| Item | Result | Probability |
|---|---|---|
| Social proof | Absent | 3% |
| Urgency | Absent | 7% |
| Opening type | Outcome promise | 93% |
| Policy flag: Guaranteed results | Check policy | 96% |
| Policy flag: Health or body claim | Clear | 2% |
| Policy flag: Income or savings claim | Clear | 5% |
| Policy flag: Addresses a personal attribute | Review | 49% |
| Policy flag: Before and after claim | Clear | 14% |
| Policy flag: Best or number one claim | Clear | 3% |
Borderline: the probability is within 0.05 of a threshold (0.70 or 0.35), so this answer can change between runs of the same text, and the score with it.
| Fact | Count | Guidance |
|---|---|---|
| Primary text | 166 characters, 27 words | Guidance 50 to 150 characters: outside |
| Headline | 18 characters, 3 words | Guidance up to 27 characters: inside |
| Words in capitals | 0 | |
| Exclamation marks | 1 | |
| Emoji | 0 | |
| Link in the text | No | |
| Price pattern | Yes |
In this sample, 4 of 7 checks pass and 1 policy flag is raised: guaranteed results. The first fix on the list is guaranteed results: Remove the guarantee, or state its real terms, such as a refund window, and read the linked policy before you run it. The counts of capitals and exclamation marks are facts, and a high count is a prompt to read the text again, not a rule. Fix the lines, run the check again and compare the two scorecards.
This sample has 1 borderline answer, so a live run of the same ad can give a different score.
The full report
After a result you can open the full report on the same page with your email: the probability and the reason for each check, a rewrite checklist in order of severity, and ready prompts filled with your ad that you paste into your own assistant. Copy it or download it as a text file. The address reaches the site owner's inbox through the form service FormSubmit. Details are in the privacy notice.
What the checker does not do
- It does not predict performance. Results depend on the audience, offer, landing page, budget and creative, and a test with real spend shows them.
- It does not approve an ad for Meta, TikTok, Google or any other platform. The platform decides, and a flag means read the linked policy page again.
- It reads text only in version 1. Images and video files are not uploaded.
- Its thresholds are the vendor's starting values, not calibrated on labelled ads.
- The model's probabilities move by a few hundredths between runs of the same text, so an answer close to a threshold can change, and the score with it. The result marks such answers as borderline.
- It does not rewrite your ad. Suggestions come from a fixed library, and the ready prompts are for your own assistant.
A checker result is a pre-flight filter and a signal. It does not predict performance, and it does not approve an ad for Meta or any other platform, because the platform decides. The thresholds are the vendor's starting values. Use the scorecard to find lines worth rewording, then test the ad with real spend.
Checker, platform review, checklist, chat assistant or paid tool
| Option | Good for | Limit |
|---|---|---|
| Platform review (for example Meta Ads Manager's own review) | The final word on policy for that platform | Runs on submit, and a rejection explains rules, not hook or offer quality |
| Manual checklist run by your team | Team process, the full creative including images | Slow, and the result depends on the reviewer |
| General chat assistant | Free-form feedback and rewrites | Answers vary with wording and run, with no fixed thresholds |
| Paid creative analytics tools | Performance data from live ads and account data | Need spend, an ad account and a subscription |
| Other free ad copy graders | A quick score for one ad | Score methods differ, so compare outputs rather than trust one |
| This checker | Pre-flight on text: typed answers, fixed thresholds, counted facts, no signup | Text only in version 1, and a signal, not a prediction |
The options do different jobs, so teams often combine two of them. Use this checker for a fast pre-flight on the words, and let the platform review and a live test have the final say.
Questions about the ad checker
- What is an ad creative checker?
- A tool that reads your ad before launch and flags weak spots. This one reads the text: hook, offer, button, audience, proof, urgency and policy risks, and counts length, capitals and exclamation marks.
- Is the ad checker really free?
- Yes. You paste your ad, you get the scorecard, and there is no account. The full report on the page asks for an email and is optional.
- Will this tell me if Meta will approve my ad?
- No. Meta decides, and its review can change after launch. The checker flags phrases that often cause trouble, such as guaranteed results, health or income claims and before and after claims, so you can read them again before you submit.
- Can it predict how my ad will perform?
- No. A check is a pre-flight filter and a signal. Performance depends on the audience, the offer, the landing page, the budget and the creative itself, and only a test shows it.
- Does it check images and videos?
- Not in version 1. It checks text. For a video, paste the hook of the first three seconds as text. Image and video files are not uploaded.
- How is it different from asking a chat assistant to review my ad?
- A chat assistant answers in prose and the answer can change with the wording. Here a decision model answers fixed yes or no and score questions with probabilities, code applies fixed thresholds, and the same status shows the same suggested fix.
- How many characters can an ad have?
- Limits depend on the placement and change over time. For Facebook Feed, Meta's ads guide recommends 50 to 150 characters of primary text and a 27 character headline (read on 15 November 2026), and the checker shows your counts next to that guidance. Confirm the placement you buy in Ads Manager, and put the hook first, since long primary text is often cut after the first lines.
- Why did the same ad get a different score?
- Each answer comes with a probability, and it moves by a few hundredths between runs of the same text. Code turns it into pass, fix or unclear at fixed thresholds (0.70 and 0.35), so an answer within 0.05 of a threshold can land on either side, and the score moves with it. The checker marks these answers as borderline. Treat a borderline answer as unclear and tighten that part of the ad.
- Is my ad text stored?
- No. Your text goes to Jev, a model run by TypeSafe AI, to score this one check. ROIMOBI does not store it, log it or send it to analytics. For the vendor's own handling of data, read TypeSafe AI's legal page.
After the check
- Get an ad account reviewed by a person
- The open-source skills behind it and user-acquisition-skills on GitHub
- Test the same ad on a synthetic audience
- Message on Telegram
The method behind the checker has a public run. In November 2026, Claude and Jev read 54 live ads, tagged them with typed answers and pre-flighted six example concepts, with a cost ledger and stated limits. See the run on 54 live ads.
Let’s make your growth add up
Tell me about your app and where growth is stuck. You can start with an audit, then move to a build and a run. Pick the step that fits.