The three steps
Draft
Paste your caption, add photos or screenshots, choose the platform and audience, and say who’s tagged. Photos are resized on your device first, so uploads are quick and location data is dropped.
Review
T.A.P. reads the draft and marks it up like a proof sheet. Each mark links a highlighted phrase, or an area of a photo, to a note in the margin. Factual claims are then checked against web sources in a second pass, and those notes fill in as the sources come back.
Ready
A short final check: have the people agreed, is anything sensitive there on purpose, have the claims been reviewed? Then you close the review and post wherever you like. T.A.P. never posts for you.
What gets checked
People & consent
Everyone involved: people you tagged or named, people visible in photos, bystanders, and whoever the post is about. For each, T.A.P. asks a specific question and offers four responses: Yes, I have permission, Ask them first (with a message you can copy), Remove / blur them, or Not applicable. It never tries to identify who someone in a photo is.
Facts
Statements are sorted into factual claims, opinions, jokes and satire, personal experience, predictions, unsupported accusations, and things that can’t really be verified. Only the claims that matter if they’re wrong are checked with sources.
Privacy
Addresses and house numbers, phone numbers, emails, usernames in screenshots, schools and workplaces, IDs and passports, boarding passes and booking codes, QR codes and barcodes, number plates, bank and order details, medical information, real-time location, and screens or documents in the background.
Context & wording
Phrasing that could accidentally accuse someone, present speculation as fact, reveal something private about another person, misquote, or leave out context that changes the meaning. Where a small rewording would help, T.A.P. suggests one while keeping your voice.
Photos & screenshots
Image-specific details, with an approximate area marked on the photo so you know what to crop or blur.
Sensitive topics
For posts about health, money, emergencies, elections, allegations or children, T.A.P. adds a short neutral note about why extra care helps.
Fact-checking and sources
Claims are checked with a live web search, so the review can reflect current reporting. Each result shows the claim, what T.A.P. found, a short evidence summary, any context that changes the picture, and the sources with publisher and date where available.
T.A.P. cannot show a citation the search didn’t return. Before a result reaches you, the server compares every source the AI cites against the list of pages the search tool actually returned. Anything that doesn’t match is removed. If that leaves a verdict with no sources, the verdict is withdrawn and the claim is marked Could not verify.
There are no truth percentages. Complicated questions aren’t reduced to a number.
What the labels mean
| Looks clear | Nothing in this area stood out. |
| Worth checking | Probably fine, but worth a second look. |
| Consent unclear | Someone is in your post and it isn’t clear they know. |
| Privacy detail | Something identifying is visible that you may not want public. |
| Needs a source | A factual claim that readers may want backed up. |
| Supported by available sources | Reliable sources found by the search agree. |
| Conflicting information | Sources contradict the claim, or each other. |
| Missing context | Roughly right, but missing something that changes its meaning. |
| Time-sensitive information | True or false depending on when; check the dates. |
| Could not verify | No reliable sources either way. Not an accusation, just honesty. |
| Opinion rather than factual claim | Yours to hold. Nothing to check. |
| Important | Used sparingly, for things like ID numbers, home addresses, a child’s details, threats, or serious allegations about a named person. |
Limits, honestly
- AI makes mistakes. It can miss things and flag things that are fine. Treat notes as a second pair of eyes, not a final ruling.
- No legal advice. T.A.P. can point out that wording makes an allegation. It can’t tell you whether something is lawful.
- Consent comes from people. T.A.P. can remind you to ask. It can’t give permission on anyone’s behalf.
- Image areas are approximate. Marked regions show roughly where something is. Check the photo yourself before you post.
- Links aren’t opened. T.A.P. considers the link as text, but doesn’t visit it.
- Video is reviewed as still frames, where available. Audio and motion aren’t analysed.