AI & its limitations

ReadRoom is powered by a large language model. We want you to use it well, which means being honest about what it can and cannot do.

What ReadRoom does

• Simulates how different audiences may interpret your words.
• Separates observation (what the text does) from inference (how it may land).
• States the assumptions it made, and its confidence.
• Offers rewrites that aim to keep your voice and your meaning.

What ReadRoom does not do

• It does not predict exact engagement, reach, or virality.
• It does not know what any specific person will actually think.
• It does not diagnose personalities or mental-health conditions, and never labels anyone.
• It does not infer protected characteristics.
• It is not a truth detector and does not verify facts.

Use your judgment

Treat every result as a second perspective, not a verdict. A bold, authentic message that a model flags as “polarising” may be exactly right for your room. ReadRoom is here to make the trade-offs visible — the call is always yours.

Configurable model

ReadRoom runs on a configurable, vision-capable Claude model. When no model is configured on a deployment, a clearly-labelled local demo engine is used so the product still works — its output is heuristic and always marked as a demo.