Inline links in AI Overviews are the real story this week

567 SISTRIX visibility points. That's what YouTube lost in Google's March core update, per Search Engine Journal's analysis of Lily Ray's Amsive data, roughly 30% larger than Wikipedia's December decline. The number bounced back, but it's the smaller news in a Pulse roundup that quietly rewrites where AI citations live on the page.

Citations are moving inline

Google announced five updates to how links appear in AI Overviews and AI Mode. The one worth your attention: inline links placed next to the sentence they support, instead of clustered at the bottom of the response.

This changes the click math. A link buried in a footer of eight other links competes with all of them and gets skipped. A link sitting next to the specific claim it backs up has a different job. It's the receipt for one sentence. If ChatGPT and Perplexity have trained users to expect cite-as-you-go formatting (they have), Google is catching up to the pattern, and the page that supplies the cited sentence wins more than it used to.

Which means the unit of optimization shrinks. It's no longer "this page is about X." It's "this paragraph is the cleanest available source for this specific factual claim." Pages that bury their numbers under three paragraphs of setup will lose to pages that put the number in the second sentence.

Forum previews change who shows up next to your name

The other update: previews from public forum discussions, surfaced alongside AI answers. Reddit, Quora, specialist forums. If your product gets discussed there, those threads now appear with your name attached.

This is uncomfortable. You don't control the forum thread. You can't edit it. The thread that ranks might be a complaint from 2022 or a comparison where a competitor wins on price. My loose read is that Google is doing this partly because forum content is the one source AI answers consistently lack good substitutes for, and partly because users keep appending "reddit" to their queries anyway, so the company may as well meet them there. I could be wrong about the second part.

Either way, the practical effect is that your brand's AI-adjacent surface area now includes threads written by strangers.

What to do this week

Pick your five highest-intent commercial queries. Run each one in Google AI Mode and in ChatGPT search. For every forum thread that appears in the previews or citations, note the thread URL, the sentiment, and whether the top comment contains a factual error about your product.

You can't rewrite Reddit, but you can answer in the thread as a verified company account, you can correct factual errors with a sourced reply, and you can flag threads where the accepted answer is two years out of date.

The core update winners-and-losers framing in Ray's analysis, what she calls "the company that owns the thing" beating "the platform to discuss it," is real for traditional search. For AI answers, both surfaces now appear in the same response. The brand site gets the inline link. The forum thread gets the preview. You need a position on both.