EN
Start free
Start free

What is grounding?

Grounding is the practice of tying an AI model's answer to specific, checkable sources, such as retrieved web pages or supplied documents, so that its statements are based on evidence instead of on the model's memory alone. It reduces invented facts.

Last updated

In more detail

An ungrounded model answers from patterns it absorbed in training, which is where invented facts come from. A grounded one is given the relevant material at the moment of asking and told to answer from it, usually showing which source supports which statement. AI search tools ground their answers in pages fetched from the web. A business can ground its own AI work by supplying the brand guidelines, the product sheet or the price list with the request. Grounding is the reason an assistant's description of your company depends on what it can find and read about you. Clear, current, consistent facts on your own site and on trusted third-party sites give it good material. The common mistake is leaving contradictory information online, such as an old address on one page and a new one on another. The model may pick either. Another mistake is assuming a grounded answer is always right; it is only as good as its sources.

Put the words into practice

Wonib learns your brand once, then makes the posts, pictures, copy, decks and documents in it, checked against the brand before you see them. Start free with 1,000 credits, no card needed.

Last updated

Start free with 1,000 credits

No card needed. Sign up with Google or your email, confirm it, and make your first piece in minutes.

Continue with Google

or

Already have an account? Sign in