Digital PR earns AI citations by getting your brand into the specific third-party publications that ChatGPT, Claude, Perplexity, and Gemini already cite in your category. AI engines build answers mostly from earned media, not your own site, so the work is finding which outlets they pull from for your topics and getting placed there. Domain Authority barely predicts the list.
Two numbers set up the whole play. Muck Rack's May 2026 study of more than 25 million cited links across ChatGPT, Claude, and Gemini found earned media drives 84% of AI citations, while a brand's own domain accounts for just 2.9%. Then Featured, analyzing 22,881 citations from 405 audits between June and August 2026, found 34.5% of those citations came from sites with Domain Authority under 40. The engines quote other people's pages, and they don't only quote the big names.
Why owned content can't win this on its own
If you optimize your own site and stop there, you're fighting for 2.9% of the citation surface. That's the Muck Rack finding, and it's held steady across three editions of their report from July 2025 through May 2026, with earned media never dropping below 82%. Journalism alone accounted for 27% of cited sources. Paid and advertorial content came in at 0.3%, so buying your way in barely registers.
This is the same shift documented in our breakdown of brand mentions vs backlinks: AI visibility is an off-page problem. Your homepage and blog still matter for the 2.9%, and they anchor your entity so an engine knows who you are, but the citations that put you inside an answer come from coverage on pages you don't control. That's what generative engine optimization rewards, and it's why PR, not just on-page work, decides who gets named.
Domain Authority is the wrong target list
Traditional digital PR ranks prospects by Domain Authority or Domain Rating. Chase the high-DR outlet, land the backlink, move on. AI citation work breaks that habit.
Featured's CEO Brett Farmiloe put it plainly: "Domain Authority tells you about traditional SEO authority. It doesn't tell you which sources AI search engines actually cite." When more than a third of AI citations come from sub-40 DA sites, a niche trade publication your category's answers already cite can be worth more than a mention in a major outlet that never surfaces for your queries. The metric that matters is not how authoritative a site looks. It's whether an engine pulls that site into answers about your topic.
That reframing changes the prospect list. You stop sorting by DR and start sorting by observed citation behavior in your specific category.
Which publications do AI engines actually cite?
There's a general answer and a specific answer, and only the specific one is useful.
The general pattern, across Semrush's three-month study of the most-cited domains and other cross-engine analyses, is that a handful of sources dominate in aggregate: Wikipedia, Reddit, YouTube, LinkedIn, and established editorial and trade press. The mix varies hard by engine. Reddit is heavily cited on some platforms and almost absent on others, and it's volatile: PromptWatch tracked Reddit's share of ChatGPT Search citations falling sharply in August 2026 after a change to how ChatGPT expands queries. Leaning your whole strategy on one platform is fragile.
The specific answer is the one you can act on. Every category has its own citation fingerprint. The publications ChatGPT cites for "best project management software" are not the ones it cites for "commercial roofing standards." You can't infer your list from a global top-50 ranking. You have to observe it for your prompts, on the engines that matter to you. Understanding how ChatGPT picks its sources helps you read the pattern, but the audit is what produces the target list.
The playbook: audit, prioritize, pitch, measure
Here's the loop. It's mechanical, which is the point.
1. Audit the citations in your category
Pick 5 to 10 prompts a buyer would actually type when researching what you sell. Run each through ChatGPT, Claude, Perplexity, and Gemini. Log two things: whether your brand appears, and which domains each engine cites in the answer. Do it by hand in a spreadsheet, or use a tool that automates it. Featured's GEO Audit runs a set of category prompts and returns which publications the four major engines cite, which competitors appear instead, and where your site is cited but your brand isn't named. It's free for one audit a day, with paid tiers at $29 and $79 per month (billed annually) as of September 2026. Other trackers in our best GEO tools roundup do versions of the same job. The output you want, however you get it, is a ranked list of the publications that keep showing up.
Read the audit for one more signal before you move on: prompts where your own page is cited but your brand name isn't in the answer. Featured flags this as a content gap, and it's a fast fix. The engine already trusts your page enough to pull it, so the problem is that your on-page copy buries the claim the answer needed. Tightening that page to state the answer up front is cheaper than earning a new placement, and it recovers a citation you nearly had.
2. Prioritize by citation frequency, not DR
Sort your audit results by how often each publication is cited for your prompts, weighted toward the engines your audience uses. A trade outlet cited in six of ten answers beats a general-interest site cited in one, even if the general site has triple the Domain Authority. Cross off publications that never appear. This is your PR target list, and it will look different from the one a traditional link-building brief would hand you.
| Target type | Typical Domain Authority | Why it matters for AI citations |
|---|---|---|
| Niche trade publication in your category | Often under 40 | Frequently cited for specific topical queries; low competition to get placed |
| Established editorial press | High | Broad trust signal; cited across many categories, harder to land |
| Community platforms (Reddit, forums) | Mixed | Heavily cited on some engines, volatile; earn genuine mentions, never spam |
| Your own blog | Varies | Anchors your entity but wins only ~2.9% of citations directly |
3. Pitch the way journalists want to be pitched
The tactics here are ordinary digital PR, aimed at a sharper list. Offer expert commentary through journalist-sourcing platforms (Connectively, Qwoted, Featured, and similar successors to the old HARO). Pitch original data, since a novel statistic gets quoted and re-quoted, which is exactly the citation multiplier you want. Keep pitches short and single-target: most journalists prefer a tight email under 200 words over a mass blast. A well-placed expert quote in one publication your category's AI answers already cite tends to outperform a pile of low-relevance guest posts.
Original data deserves its own line, because it's the highest-leverage pitch you can send. A survey, a benchmark, or an analysis of your own operational numbers gives a journalist something to cite and gives every downstream writer a stat to reuse. Each reuse is another page an engine might pull. You don't need a large study. A single defensible number that nobody else has published can seed months of coverage, which is why data-led PR consistently outperforms opinion pitches for this specific goal.
Two details specific to AI citations. First, an unlinked mention still counts, because engines read what the web says about you, not just who links to you, so don't kill a placement over a nofollow. Second, get named clearly and consistently. The engine has to resolve every mention to the same entity, so use one brand name and description everywhere.
4. Measure citation share, then repeat
Re-run the same prompt set on a schedule and watch whether you start appearing and which of your placements get cited. This is a share-of-voice metric, not a rankings check. Fold it into a single LLM optimization program rather than treating it as a one-off campaign. Expect a quarter of consistent pitching before share moves meaningfully, since coverage has to be published, crawled, and pulled into retrieval first.
Where this fits, and where it doesn't
Digital PR for AI citations is not the same as chasing a licensing deal. Some publishers have signed content agreements with AI companies, and that's a separate advantage we cover in AI licensing deals and citations; most brands can't buy a seat at that table and don't need to. Earning coverage on the publications engines already cite is the path that's open to everyone.
It also doesn't replace on-page work. You still need pages structured the way answer engines reward, so that when an engine does reach your site, it can extract a clean answer. Think of it as two fronts: PR earns the third-party citations that carry the weight, and answer-engine formatting captures the smaller share of direct citations plus the traffic that follows.
The strategic mistake is spending the whole budget on your own content and a generic high-DR link campaign, then wondering why AI never names you. The data points the other way. Find the publications the engines cite in your category, earn your way onto them, and measure the share you win. That's the job.
- https://www.globenewswire.com/news-release/2026/08/25/3350347/0/en/featured-launches-geo-audit-to-show-brands-which-publications-ai-search-engines-cite-and-which-ones-to-pitch.html
- https://muckrack.com/blog/what-is-ai-reading-may-2026
- https://www.globenewswire.com/news-release/2026/05/07/3290268/0/en/generative-pulse-earned-media-consistently-drives-ai-citations-holding-at-84.html
- https://www.semrush.com/blog/most-cited-domains-ai/
- https://www.promptwatch.com/data/reddit-citations-are-dropping-in-chatgpt
- https://backlinko.com/digital-pr-strategies
