How to Get Cited by Perplexity, Claude, and Copilot: What the Data Shows

ChatGPT and Google AI Overviews get most of the GEO attention, but Promptwatch's dataset covers Perplexity, Claude, and Microsoft Copilot too, and the three behave nothing like each other. One is the most predictable engine tracked. One is a citation crawler that barely existed nine months ago and is now growing 100x. One is still too unstable to plan a strategy around. Here's what the data says to do with each.

Perplexity: the most predictable engine, and the best GEO test bench

Promptwatch's citation-count tracking calls Perplexity "the most predictable engine in this report": almost exactly ten sources per answer, day after day, with barely a decimal of movement. That consistency points to a deliberate retrieval design rather than emergent behavior, Perplexity is built to show its research, and the citation count doesn't drift with model updates the way ChatGPT's and Copilot's do.

That stability makes Perplexity useful in a way the other engines aren't: because its citation count barely moves on its own, a change in your citation pickup there reflects your content changes, not engine noise. If you want to test whether a rewritten answer block, a new comparison table, or a schema fix actually moves citations, run the experiment on Perplexity first.

Perplexity's social and LinkedIn behavior also diverges from ChatGPT's. On social citations, Perplexity leads with YouTube (2.67%), not Reddit, closer to AI Overviews' pattern than ChatGPT's. And on LinkedIn, Perplexity is the one engine where ordinary feed posts (41.88%) beat polished Pulse articles (32.46%), its retrieval clearly favors fresh, specific, practitioner takes over formal long-form writing.

What to do with this:

Use Perplexity as your GEO test bench. Ten reliable slots and low noise mean you can actually measure whether an edit worked.
If Perplexity matters for your audience, invest in YouTube alongside written content, it outperforms every social platform there except Reddit.
Post to the LinkedIn feed directly about timely, specific topics; a strong post is worth more than a polished article on this one engine specifically.

Claude: still small, but growing 100x in four months

Claude doesn't show up in most citation-share reports yet because it's a small piece of the overall pie, but Promptwatch's crawler data shows why you shouldn't ignore it. Claude's citation crawler (Claude-User) went from roughly 30 visits a day (~0.04% of tracked AI citation crawler traffic) in mid-December 2025 to several thousand a day by mid-April 2026, peaking at 1.73% daily share on April 13, a more than 100x increase in four months. The inflection point was late March, when daily visits jumped from the hundreds into the thousands within two weeks.

Claude is still around 1% of tracked citation crawler traffic overall, with ChatGPT dominating the rest, this is a fast-growing secondary channel, not where your optimization budget should concentrate today. But the growth curve means the footprint will keep expanding whether or not you do anything, and the only real risk is being invisible to it by accident.

What to do with this:

Check that Anthropic's crawlers (Claude-User, ClaudeBot, anthropic-ai) aren't blocked in robots.txt or by CDN/WAF bot rules. A block that was harmless in December now silently excludes you from a channel that's growing every month.
Search your server or CDN logs for Anthropic user agents to establish a baseline, then review it monthly, the growth curve suggests your own numbers will keep moving too.
Treat Claude as a channel to unblock and monitor now, and to actively optimize for once its share grows past the low single digits.

Copilot: too volatile to be a primary target right now

Promptwatch's citation-count data calls Copilot "the wildcard of the four engines tracked": its average citation count has swung from under two sources per response to nearly seventeen within a few weeks, before settling at the low end recently. Swings that large mean Microsoft is still actively re-architecting how Copilot retrieves and attributes sources, when the average sits low, Copilot offers the fewest citation slots of any engine tracked, scarcer even than ChatGPT.

What to do with this:

Make sure Bing indexes and ranks your key pages well. Copilot draws on Microsoft's search stack, and with citation slots currently scarce, only well-ranked content gets pulled in.
Judge Copilot performance on monthly trends, not weekly snapshots, a sudden swing in your own citations is more likely a retrieval change on Microsoft's side than something you did.
Use Bing's free AI Performance report and submit changes via IndexNow, but keep Copilot on your radar rather than at the center of strategy until its citation behavior stabilizes.

Universal patterns that carry across all three (and every engine)

Two Promptwatch reports cut across Perplexity, Claude's broader ecosystem behavior, and every other engine, and both reward the same unglamorous habits:

Reddit citations skew old and unpopular. ~71% of cited Reddit posts had fewer than 10 upvotes, and ~60% of all cited posts were more than six months old. AI retrieval selects for topical relevance to the exact question asked, not popularity, an obscure thread that precisely answers a niche question outperforms a viral one that only grazes it.
YouTube citations skew the same way. ~80% of cited videos had under 100K views and under 100K subscribers; 54% of cited videos were over two years old. Small, focused channels that consistently cover one topic area get picked up even without a large audience.

What to do with this:

Stop optimizing for upvotes, likes, or subscriber counts as a proxy for AI citation odds, the data says they barely correlate. Optimize for a single, complete, correctly-worded answer to a specific question instead.
Answer new and low-competition threads in your niche communities; a helpful reply with 5 upvotes has roughly the same citation odds as one with 500.
Treat your Reddit and YouTube contributions as a compounding, multi-year asset. What you publish now is disproportionately likely to be what's still being cited in 2027 to 2028.

Where this leaves Gemini

Gemini doesn't have a dedicated standalone citation-share report in Promptwatch's public dataset the way ChatGPT and Google AI Overviews do, because Gemini's consumer search behavior largely surfaces through Google AI Overviews and AI Mode, which are covered in depth separately. The average-sources-per-response tracker above includes Gemini among the four engines it monitors for citation-count behavior, so the same "measure separately, don't assume one engine predicts another" rule applies here too.

FAQ

Which of these three engines should I prioritize?

Perplexity if you want a stable, measurable test bench and your audience skews toward researchers and early adopters. Claude if you want to be ahead of a fast-growing channel before it matters at scale. Copilot only if Windows/Edge/Office reach is already a meaningful part of your funnel, its citation behavior hasn't stabilized enough to optimize against confidently yet.

Is Claude worth optimizing for today?

Not as a primary channel, it's still roughly 1% of tracked citation crawler traffic. But it's worth unblocking and monitoring today, since a 100x growth curve in four months means today's minor channel can become tomorrow's meaningful one, and being blocked from the start is a much harder position to recover from.

Why does Copilot's citation count swing so much?

Promptwatch's data doesn't attribute a specific cause, only that the swings (2 to 17 sources per response) are large enough to indicate Microsoft is still actively changing how Copilot retrieves and attributes sources. Plan against monthly trends, not any single snapshot.

Do likes and upvotes help get content cited on these engines?

No, based on the Reddit and YouTube data above. Citation selection tracks topical precision to the exact question asked, not popularity or engagement.

What to do this week

Search your logs for Anthropic crawler user agents (Claude-User, ClaudeBot) and confirm they're not blocked.
Run 10 to 15 of your target prompts on Perplexity specifically and record the citation baseline, it's your most reliable engine for measuring whether a content edit worked.
Answer two or three specific, low-competition questions in niche subreddits or a focused YouTube video rather than chasing views or upvotes.
Check Bing Webmaster Tools' AI Performance report if Copilot is a channel you care about, and confirm IndexNow is set up.
Track all of it in one place. Promptwatch Essential covers ChatGPT, Perplexity, Gemini, Claude, Copilot, and Google AI Overviews daily with citation analytics for $95/mo. Professional adds 25M AI crawler logs at $245/mo. That gives you one dashboard instead of five separate spot checks.