Interactive moderation demo
See exactly what reaches the screen
QR Wall moderates a live event wall in two independent layers on the Professional plan. AI screening rates every message and photo from 1 (clean) to 5 (blocked) and keeps anything above the strictness level you choose off the screen. Manual approval holds every post in a queue until you or your team approve it from a phone or laptop. You can use either one or both together. Below, 22 posts of the kind a junior sports meet receives play through both layers, with hold on, the way youth sport and agency events run it: approve or block each post, change the level, or switch hold off to see what AI screening does alone.
Last updated: October 2026
Every post, at every level
The ratings below are for AI screening alone, with hold switched off. A post shows when its rating is at or below the level the host picks. Strictness levels, each allowing everything below it too: 1 · Professional only; 2 · Casual & enthusiastic; 3 · Fun & mild slang; 4 · Strong language; 5 · Anything legal. Ratings are the expected classifier output for each post. The two stand-in photos borrow the rating of the matching text post; we do not publish explicit or hate imagery.
| Post | AI rating | Level 1 | Level 2 | Level 3 | Level 4 | Level 5 | Why |
|---|---|---|---|---|---|---|---|
| “Go Tilly!! Nailed that beam routine 🤸♀️” | 1 (clean) | Shows | Shows | Shows | Shows | Shows | A parent cheering. Clean at every level. |
| [Photo: A gymnast on the balance beam at a competition] | 1 (clean) | Shows | Shows | Shows | Shows | Shows | Clean. The AI does not know, or check, whether the person in a photo is a child. See "Kids in photos" below. |
| “LET'S GOOOO BLUES!!! 💙💙💙” | 2 (mild) | Held back | Shows | Shows | Shows | Shows | Shouting and emoji read as enthusiastic: level 2. |
| [Photo: Fans waving flags in a stadium] | 1 (clean) | Shows | Shows | Shows | Shows | Shows | A crowd shot. Clean. |
| “Ref is blind 🙈” | 3 (edgy) | Held back | Held back | Shows | Shows | Shows | Banter. Not harmful, so the harm model alone would pass it; the tone check rates it edgy. Shows from level 3. |
| “Proud of you Mia, see you at the canteen after 😂” | 1 (clean) | Shows | Shows | Shows | Shows | Shows | Clean. |
| [Photo: Friends clinking glasses of beer] | 2 (mild) | Held back | Shows | Shows | Shows | Shows | Alcohol, rated casual. The AI does not know your event is a junior meet. That judgement is yours, in the queue. |
| “Booo, the red team cheated” | 3 (edgy) | Held back | Held back | Shows | Shows | Shows | Not harmful, but not friendly either: edgy. Hidden at levels 1–2. |
| “Damn that vault was insane” | 3 (edgy) | Held back | Held back | Shows | Shows | Shows | Mild slang. Shows from level 3 ("Fun & mild slang"). |
| [Photo: A young gymnast training on the beam] | 1 (clean) | Shows | Shows | Shows | Shows | Shows | Clean. |
| “This team sucks lol” | 3 (edgy) | Held back | Held back | Shows | Shows | Shows | The classic fan-wall post: no harm category fires. Only the tone check, or a person, stops it. |
| “That judge is a f*cking joke” | 4 (explicit) | Held back | Held back | Held back | Shows | Shows | Disguised swearing. The profanity list sees through the asterisk and sets the floor at 4. |
| [Photo: A man making a rude hand gesture] | 4 (explicit) | Held back | Held back | Held back | Shows | Shows | A rude gesture. The harm model does not score gestures; the tone check does: explicit, 4. |
| “dig bick energy from the coach 😂” | 4 (explicit) | Held back | Held back | Held back | Shows | Shows | Wordplay built to slip past word lists. Caught by the tone check, not the list. |
| “Puta madre, qué salto” | 4 (explicit) | Held back | Held back | Held back | Shows | Shows | Spanish swearing. The word list is English; the AI layers catch it. |
| [Photo: A gymnast with a ribbon] | 1 (clean) | Shows | Shows | Shows | Shows | Shows | Clean. |
| “Call me 0412 345 678 😉” | 1 (clean) | Shows | Shows | Shows | Shows | Shows | A phone number. Nothing harmful, so the AI passes it at every level. Only a person in the queue catches this. |
| “Go back to where you came from” | 5 (blocked) | Held back | Held back | Held back | Held back | Shows | Harassment aimed at a group. Blocked at every level, including 5. |
| [Hate-symbol photo, stand-in] | 5 (blocked) | Held back | Held back | Held back | Held back | Shows | Stands in for a hate-symbol photo, which we will not put on this page. Hate figures and symbols are a hard 5 in the tone check. |
| “Send nudes 😏” | 5 (blocked) | Held back | Held back | Held back | Held back | Shows | Sexual solicitation. Blocked at every level. |
| [Explicit photo, stand-in] | 5 (blocked) | Held back | Held back | Held back | Held back | Shows | Stands in for an explicit photo, which we will not put on this page. Nudity is blocked at every level. |
| [Photo: A packed arena crowd] | 1 (clean) | Shows | Shows | Shows | Shows | Shows | Clean. |
What the AI checks
A harm model
OpenAI's moderation model reads the text and the photo together and scores 13 harm categories: sexual content, hate, harassment, violence, self-harm and more. The highest score sets the category, from 1 to 5.
A tone check
A second AI asks a different question: is this fine to show at this level? It catches what harm models miss: "this team sucks", insults disguised as wordplay, rude gestures, and memes that reference hate figures.
A swear-word list
An English profanity list that sees through spacing and asterisks. A match sets the post to at least 4, so it never shows below "Strong language".
Video clips are checked frame by frame: up to five frames, one every four seconds, each screened like a photo. The worst frame decides the whole clip. Clips are muted, so there is no audio to check.
Kids in photos
The AI does not detect whether a child is in a photo, and it does not try to. A clean photo of a young gymnast is rated clean, because it is. For a youth event the question is not "is this harmful?" but "should this child be on a screen?", and only a person can answer that.
For junior events, turn on Hold everything for approval so a person sees every photo before the room does. Add Camera-only mode, which stops guests uploading old photos or screenshots from their camera roll, and the public-display consent tick, which asks everyone who posts to agree that their photo or message may be shown on the event screen.
Questions buyers ask
Can I moderate an event photo wall before posts appear?
Yes. On the Professional plan, turn on "Hold everything for approval" and every message and photo waits in a queue until you or your team approve it from a phone or laptop. Reactions skip the queue because they can only be the eight you chose.
What does AI moderation catch on a live photo wall?
Nudity and sexual content, hate and harassment, violence, swearing (including disguised spellings), rude gestures and hate-symbol memes. Each post is rated 1 to 5 and you choose the highest level that may show. It does not catch personal details such as a phone number, and it cannot judge context, such as alcohol at a junior event.
Is AI moderation enough for a kids or school event?
No, and we would not claim it is. The AI does not detect minors. For any event with children, use manual approval (and optionally camera-only mode and the consent tick) so a person sees every photo before it reaches the screen.
What happens to a post the AI blocks?
It never reaches the wall, but it is kept: it waits in an "AI blocked" tab where you can let it through if the AI was wrong. The guest is not told it was blocked.
How many moderators can work the queue?
The queue opens from the wall owner's account on any number of phones or laptops at once. Approving or blocking the same post twice is harmless.
Which plan includes moderation?
Both AI screening and manual approval are on the Professional plan, along with your logo in the QR code, your own background and camera-only mode.
Use this on your event
AI screening and manual approval are on the Professional plan. Open this page in a browser to play the demo: guests' posts arrive on a phone, you approve or block them in the real queue, and the approved ones land on a live wall.
Use this on your event →