How to Track AI Search Citations for Press Releases

A finance team at a Singapore-listed company issues its quarterly update at nine in the morning. By lunchtime, someone on the communications side opens ChatGPT and asks what the company just announced, half expecting silence and half hoping to see their own release staring back. That small, slightly anxious test is really a question about AI search citations: whether anyone’s chatbot is quoting what the company actually wrote, and how a PR team would even know.

AI answer engines have moved from a curiosity to a measurable source of traffic, even if that measure is still small. Communications teams who spent years tracking media pickups and backlinks are starting to ask a newer question: is a chatbot actually citing what we published, what would count as proof, and whether that proof belongs anywhere near a monthly PR report.

How Much Traffic AI Search Actually Sends

The honest answer, at least for now, is not very much. A 2026 study tracking referral data from more than 101,000 websites found that AI search engines accounted for roughly 0.32 percent of all website traffic in early 2026, up from 0.02 percent in 2024, a sixteenfold increase in relative share even though organic search still sends over a hundred times as much traffic (SE Ranking’s AI traffic study). ChatGPT alone accounts for around three-quarters of that AI referral traffic, with Google’s Gemini and Anthropic’s Claude both growing fast off a small base, and treating AI citation tracking as PR’s new scoreboard would still be getting ahead of the data.

Why Some Releases Get Cited and Others Don’t

Not every kind of query treats a press release the same way. One widely discussed AI-citation tracking experiment distributed two press releases through a wire service and then monitored citations across five major AI engines (an AI-visibility case study from Otterly.ai). Within two days, Microsoft Copilot had already cited the release, and within a week ChatGPT, Google’s AI Overviews, Perplexity and Gemini had all picked it up too, well before the story had finished circulating through the usual editorial channels.

That coverage drew heavily on the finance and news-aggregator sites the release had syndicated to, rather than the newswire’s own page alone. The experiment covered one brand over two weeks, so it should be read as one illustration, not a universal rule. Even so, the pattern lines up with how these systems are generally understood to behave: a fresh, clearly dated announcement tied to a named company is the kind of query an AI engine can often answer directly from a press release.

Broader “best X” or category-style questions behave quite differently, and tend to pull from editorial comparison sites, review platforms and listicles, because those sources are built to rank and compare, not to announce news. For a Singapore firm issuing a product launch or a funding update, the realistic opportunity sits with brand-name and news-style queries, not generic category searches, and setting that expectation early saves a team from chasing the wrong kind of visibility.

Four Places to Look for Evidence

Checking whether a release has been cited does not require an expensive platform, at least to start. A few habits cover most of what a communications team needs, and a dedicated AI-visibility monitoring tool only starts to earn its cost once a team is running these checks across dozens of releases each quarter instead of one or two:

• Prompt the major engines directly with brand and product-name queries a few days after each release goes out, and note which ones surface a citation or a working link.
• Check Google Analytics 4 referral sources for traffic arriving from chatgpt.com, perplexity.ai, gemini.google.com or copilot.microsoft.com, which most GA4 setups now group separately from standard organic search.
• Ask the distribution partner whether their reporting includes AI-engine pickup alongside the usual outlet list, since several wire services have started adding this.
• Revisit the same prompts a week or two later, since citations in these systems can appear, disappear and reappear as models refresh what they retrieve from.

What Actually Improves the Odds

Citation tracking only tells a team what already happened; it does not, on its own, change the outcome. We’ve already covered how schema markup for press releases gives AI systems a cleaner structural read of who, what and when, even though the evidence that markup alone drives citation is still thin, and the two tactics work best treated as separate projects with separate expectations.

Syndication reach counts for as much as the writing itself: a release picked up by a recognised finance or news-aggregator site appears to carry more weight with these systems than an identical release left on a single domain. Clear, specific headlines that name the company and the news tend to survive the content filters these engines apply before they quote a source, more reliably than vague marketing language does, and that discipline pays off in ordinary search results too.

None of this replaces the fundamentals of a well-targeted distribution plan or a genuine relationship with the right journalist, and no amount of AI-citation tracking will rescue a release with a weak story at its centre. A results announcement still has its own SGX disclosure timeline to meet first, which is a compliance matter entirely separate from anything covered here and should be confirmed with a company’s own advisers. As AI answer engines keep growing their share of search traffic, tracking whether a release actually gets cited is becoming a reasonable line to add to a PR report, alongside pickups and backlinks.

If your team would like a second pair of eyes on where your releases are landing, get in touch with PressRelease.com.sg and we will walk through what we are seeing.