# AI Grounding

## Quick answer

AI grounding is the step where an AI answer engine ties its generated response to specific outside sources, usually shown as citations or links, instead of answering purely from what the underlying model learned during training. A grounded answer can be traced back to the pages it drew on; an ungrounded one cannot.

## Why it matters

When an engine's answer about a brand is grounded, the cited source is the direct, fixable cause of what the answer says, whether that is an outdated page, a competitor's comparison article, or the brand's own content. When an answer is not grounded and comes from the model's general training instead, there is no specific page to fix, only the slower work of waiting for the brand's newer content to eventually show up in a future training run. This is why two answers that sound equally wrong can call for very different responses depending on whether either one shows its sources.

## How to measure it

Check whether an AI engine's answer about the brand includes visible citations or sources, and if it does, read what those sources actually say. An answer with no citations at all is a different problem than an answer with citations pointing at the wrong or outdated page, and the two call for different fixes: getting the brand into the sources an engine already reads, versus correcting what a specific already-cited source says. Repeating this check across engines also shows which ones tend to ground their answers in sources at all, since not every engine does so by default.

## Example

A password manager asks why an AI engine describes its free plan incorrectly, and finds the engine's answer cites a two-year-old blog post rather than the vendor's own current pricing page. Because the answer was grounded in a specific, identifiable source, the team has a direct target, getting that source updated or outranked by a current one, rather than guessing at what the model might have memorized. A second engine answering the same question shows no citations at all, which the team treats as a separate, harder problem to chase.

## FAQ

### Is every AI answer grounded?

No. Some answers come with visible sources and some do not; an answer engine like Perplexity shows sources by default, while a general chat answer may rely more on what the model already knows.

### Why does grounding matter beyond simply getting mentioned?

Because a grounded mention points at a specific source a brand can actually check and try to change, while an ungrounded mention offers no clear next step.

### Do all AI engines show their grounding sources the same way?

No. Some, like Perplexity, show sources by default on most answers; others show citations only sometimes or not at all, depending on the question and the engine's own design.

## Related terms

- /glossary/retrieval-augmented-generation
- /glossary/perplexity
- /glossary/ai-visibility

## Related

- AI citation tracking: /blog/ai-citation-tracking