90%
What it means
An ungrounded answer comes from the model's memory. A grounded answer is tied to sources: the AI searches, reads, and writes from what it found, usually with links.
Google uses the word for Gemini's "grounding with Google Search". It's the same idea as RAG.
Why it matters
Grounding reduces made-up facts and lets AI answer about recent things. For brands, it means the pages AI finds today decide what it says, more than its training.
That's good news: you can change those pages, or get onto them, now.
What our data shows
Grounded answers lean on lists. In our study, 90% of the brands AI recommended were on a "best X" list the engine read for that answer.
Example
Gemini is asked "is Acme CRM good for small teams?". Without grounding, it gives a vague answer. With grounding, it searches, reads a recent review and Acme's pricing page, and answers with today's price and a summary of the review, with links.
Frequently asked questions
Is grounding always on?
Not always. AI tools decide when to search. Buying and current-events questions usually trigger it.
Does grounding stop hallucinations?
It reduces them. The AI can still misread a page, or read an out-of-date one.
Can I control what sources an AI grounds on?
Not directly. You can make your pages crawlable and clear, and get onto the pages it already reads.