Sources on show: designing AI citations people actually check

AI answers increasingly come with sources: small numbers, links or cards that point to where the information came from. In principle, that lets people check. In practice, many citation designs make checking so awkward that hardly anyone does.

Why sources matter#

AI systems can state wrong things with complete confidence. Sources are the main way people can tell a well-supported claim from a shaky one. If the design discourages checking, the sources become decoration, and people either trust everything or nothing.

Link claims, not just answers#

A list of five links at the end of a long answer doesn’t tell people which link supports which statement. Placing a small marker next to each claim, linked to its source, makes checking one specific fact quick.

Show the source before the click#

People decide whether a source is worth opening from its name and type. Show the publication or site name, the page title and, ideally, the date. A marker that reveals nothing until clicked asks people to click blindly, and most won’t.

Show the relevant part#

Opening a long page to hunt for one sentence is a chore. Where possible, show the passage that supports the claim, or jump straight to it.

Be honest about gaps#

When a claim has no good source, say so. “We couldn’t find a reliable source for this” is more useful, and more trustworthy, than a weak link dressed up as support.

What this means for your website#

If AI assistants cite your pages, the same principles work in your favor. Pages that state facts clearly, with a visible date, author and organization name, are easier to quote and easier for people to trust when they arrive. See also SEO when AI answers first.

A quick checklist#

• Are sources linked to individual claims?
• Can people see the source’s name and type before clicking?
• Does opening a source lead straight to the relevant part?
• Are unsupported claims clearly marked?