Getting Cited by AI: Why This Is a PR Problem, Not Just an SEO One

Something changed in how buyers find companies, and the industry response has been to rename SEO. Generative engine optimization, answer engine optimization, LLM SEO, AI search optimization: four labels for a set of tactics that, as I see it, mostly assume the old mechanism still applies, just with a new crawler.

I don't think it does, and the public data on what these systems cite points the other way. If you're asking how to get cited by AI, my answer is that it depends mainly on what other people write about you, with your own pages as the smaller lever: in Muck Rack's 2026 tracking, more than 84% of AI citations came from journalism and other third-party sources, and about 15% from companies' own sites.

A note on that data before using it. Almost everyone publishing numbers on AI citations sells AI-visibility tools or services, including Ahrefs, Muck Rack, Adobe and Seer Interactive, all cited below. The one academic study here thanks a GEO company for its support. Where their findings line up, they're worth taking seriously.

Key takeaways

• In University of Toronto tests on U.S. software questions, about 73% of the sources ChatGPT's search model cited were earned media, meaning third-party publishers and reviewers, against about 45% of Google's top results.
• In the same researchers' August 2025 tests on car and consumer-electronics questions, fewer than one in ten of the domains cited by the search versions of ChatGPT or Perplexity were cited by both.
• In Seer Interactive's full-year 2025 data on 53 brands, brands labeled as cited in a Google AI Overview on informational queries had a higher organic click-through rate than uncited brands: 2.07% against 1.61%, once one account that made up 47% of the uncited impressions is excluded. Citation status came from one recent snapshot.

The two systems barely overlap

When University of Toronto researchers ran the same car and consumer-electronics questions through the search versions of ChatGPT and Perplexity in August 2025, using the companies' developer APIs, fewer than one in ten of the domains either engine cited were cited by both. To me that looks less like two search engines with different rankings and more like two different ways of choosing sources.

And ChatGPT doesn't always search. OpenAI says it searches the web when a question "would benefit from current information." In Ahrefs' sample of 66.6 million responses to free, logged-out users in August and September 2026, only about one in seven contained the search queries ChatGPT runs when it looks something up. Questions comparing products trigger searches more often, and Ahrefs cautions that a response without visible search queries isn't proof nothing was looked up. Still, my working assumption is that much of what ChatGPT says about your category isn't a fresh look at your site, and that a homepage edit today won't reach those answers.

Why is getting cited by AI a PR problem?

If an AI system answers from accumulated knowledge rather than a live crawl, you don't get into the answer by editing your site. My bet is that you get in the way reputations have always been built: by being written about, discussed and referenced widely enough that the model absorbed you as part of your category. No citation study can prove that recipe, but I don't know a better one.

When these systems do search, the pattern points the same way. In the Toronto tests, about 73% of the sources ChatGPT's search model cited for U.S. software questions were earned media, meaning third-party publishers and reviewers, against about 45% of Google's top results for the same questions. Muck Rack's tracking across ChatGPT, Claude, Gemini and Perplexity puts journalism, reference sites, research and other third-party sources at more than 84% of AI citations in 2026. A company's own content was about 15%.

Not all of that is press coverage. Muck Rack also counts reference sites, research, government pages and other companies' content. Some of those sources supply background without mentioning any company, and a share of citations doesn't measure how much more coverage would help you. Still, what other people publish makes up far more of what these systems cite than what companies publish about themselves. That is how reputation has always worked, arriving through a new door.

My read of the tactical implication, uncomfortable for anyone selling AI-visibility software, is that the highest-leverage work is third-party coverage, credible mentions, presence in the places these systems cite most, and being described consistently across all of them. Schema markup, crawler access and content structure still matter, and when ChatGPT does search, Ahrefs found most of the pages it cites come through ordinary search results, so your pages still have to rank. That work pays off mainly when an assistant searches and finds you. For answers a model gives from what it already knows, your pages and your coverage can both end up in future training data, neither changes a model that's already trained, and no citation study yet shows which counts for more. My bet, as above, is on coverage.

Is AI search traffic worth the effort?

AI assistants still send a small share of website visits. Across the 118,671 sites in Ahrefs' traffic benchmark, AI assistants sent about 0.5% of visits in October 2026, against about 28% from search engines. ChatGPT alone sent about four in five of those AI visits.

The conversion isn't small, though the best data so far comes from retail. In Adobe's analysis of more than a trillion visits to U.S. retail sites, visitors arriving from AI sources converted 54% better than other traffic in May 2026, a reversal from a year earlier. B2B data is thinner, and much of it comes from vendors measuring their own sites, so don't treat a retail number as your benchmark; track leads from AI referrals in your own analytics. My guess at why the visits are worth more: someone arriving from an AI answer has often already had your company named to them by a system they chose to ask, which is a warmer introduction than a blue link.

On Google, being cited in an AI Overview goes with more clicks, not fewer. Seer Interactive tracked 53 brands from January 2025 to February 2026. On informational queries showing an AI Overview, brands cited in it got more organic clicks per impression than brands that weren't. Seer's headline figure is 120% more, but one account made up 47% of the uncited impressions; without it, Seer's own numbers put cited brands at 2.07% click-through against 1.61%, nearer 30% more. Those are full-year 2025 averages, and Seer labeled each query cited or uncited from its most recent results snapshot, applied to every month, so the comparison isn't month by month. Cited brands still got fewer clicks than on queries with no overview at all, and Seer says it can't prove the citation caused the difference. Losing clicks to the overview is real. Being left out of it looks worse.

How do you get cited by AI?

Getting cited by AI takes work on and off your own site, and in my view the off-site work matters more.

Get covered by the sources these systems cite. For ChatGPT that means broad, consistent third-party mentions built over time, starting with the reporters who cover your category. Perplexity, in the Toronto tests, drew more on brand and social content than ChatGPT did, so I'd give community discussion more weight there, after checking which sources each platform cites in your category today. These are mostly earned-media problems with different targets, and I wouldn't run the same play for both.

Be described the same way everywhere. If your company is a "developer platform" in one place, an "AI infrastructure company" in another and a "workflow automation tool" in a third, you've given buyers, reporters and the systems summarizing them three answers instead of one. I think consistency across coverage matters more here than clever phrasing anywhere.

Publish things that are quotable and checkable. Original data, clear definitions, specific numbers with sources. A page full of positioning language gives a writer, or a model, nothing to lift.

Keep something recent in circulation. In Ahrefs' study of 17 million citations, the pages AI assistants cited were on average about 26% newer than the pages in Google's organic results, and ChatGPT leaned newest of all. The average cited page was still nearly three years old, so older coverage keeps working. I just wouldn't let six months go by with nothing new.

Answer real questions directly. Build pages around the actual question a buyer types and answer it in the first paragraph. In Ahrefs' study of 1.4 million ChatGPT prompts, the pages it cited had titles measurably closer to the user's question than the pages it retrieved and passed over. When we rebuilt our own site, I cared far less about how many blog posts we had than about whether each one answers something a buyer actually asks.

What to stop doing

Stop treating this as a technical checklist. In my view the tooling market has moved faster than the evidence, and much of what I see sold as AI-visibility optimization is conventional on-page SEO with new labels attached. Ask any vendor to tie its recommendations to measured changes in visibility or qualified traffic.

Stop measuring it like search. There's no rank position, and a single check proves little: in Muck Rack's tests, ChatGPT asked the same question twice in one day repeated only 41% of the domains it cited (Perplexity repeated 94%). The useful measurement is whether you appear in answers to the questions your buyers actually ask, checked by asking them, on each platform separately, on a schedule.

When we run that check, we read the answers behind the count, not just the count. A tally of citations can include questions that have nothing to do with what a company sells. A citation count tells you little until you know which questions it answers.

Common questions

How do you get cited by AI? In my view, mainly by being written about. In Muck Rack's 2026 tracking across ChatGPT, Claude, Gemini and Perplexity, more than 84% of AI citations came from journalism and other third-party sources, and about 15% from companies' own sites. Coverage in the outlets your category's AI answers draw on matters most, alongside consistent descriptions and pages that rank and answer buyers' questions directly.

What is LLM SEO? The practice of getting a company or its content named and cited in answers from systems like ChatGPT, Perplexity and Google's AI Overviews. The name is misleading, because in my view the bigger lever is what third parties publish about you, alongside useful pages and sound technical SEO.

How do you rank in ChatGPT? There's no fixed position to earn. ChatGPT doesn't search for every answer, and when it does, most of the pages it cites come through ordinary search results. So the work is relevant third-party coverage, useful pages that rank, and checking whether you're named across repeated buyer questions.

Do ChatGPT and Perplexity cite the same sources? Mostly not, in the evidence so far. In University of Toronto tests run through the engines' developer APIs in August 2025, on car and consumer-electronics questions, fewer than one in ten of the domains either one cited were cited by both. Results may differ today, so check each platform separately.

Is AI search traffic worth pursuing when volume is low? It can be, but judge it by the qualified leads it brings, not the volume. AI assistants sent about 0.5% of visits across the sites in Ahrefs' October 2026 benchmark, but in Adobe's retail data, visitors from AI sources converted 54% better than other traffic. B2B data is thinner, so measure your own.

Sources

• Mahe Chen, Xiaoxuan Wang, Kaiwen Chen and Nick Koudas, Generative Engine Optimization: How to Dominate AI Search, arXiv (University of Toronto), September 10, 2025
• Matt Albasi, New report: Earned media still drives more than 84% of AI citations, but the sources behind AI answers are shifting, Muck Rack, October 28, 2026
• OpenAI, Searching the web with ChatGPT, OpenAI Help Center, accessed November 8, 2026
• Louise Linehan, How ChatGPT Does Query Fan-Out (15 Million Queries Studied), Ahrefs, October 12, 2026
• Ahrefs, AI vs Search Traffic Analysis, updated November 7, 2026
• Abbas Haleem, Adobe: AI-referred traffic to retail sites doubles in a year, Digital Commerce 360, June 17, 2026
• Tracy McDonald, Hannah Cooley and Marketa Williams, AIO Impact on Google CTR: 2026 Update, Seer Interactive, April 24, 2026
• Ryan Law, New Study: AI Assistants Prefer to Cite "Fresher" Content (17 Million Citations Analyzed), Ahrefs, July 28, 2025
• Louise Linehan, Why ChatGPT Cites One Page Over Another (Study of 1.4M Prompts), Ahrefs, April 15, 2026

Salient PR runs earned-media and answer engine programs for AI and technology companies. Citation and traffic figures reflect the published studies listed above, as of November 2026.