TL;DR
- Digital PR earns the third-party mentions AI engines cite.
- Earned media drives 84% of AI citations, paid content 0.3%.
- Target what each engine cites: trade press, Wikipedia, Reddit, YouTube.
- Data studies, expert quotes, and product-source pages earn the coverage.
- Measure pooled citation share per engine, never one answer.
Digital PR is the practice of earning brand mentions, editorial coverage, and citations on websites you do not own. In 2026 that coverage has a second audience: AI engines. Muck Rack’s May 2026 analysis of more than 25 million links cited by ChatGPT, Claude, and Gemini found that earned media accounts for 84% of all AI citations. Paid and advertorial content accounts for 0.3%. The coverage a digital PR campaign earns is now the source material of AI answers.
That changes who digital PR targets, what campaigns produce, and how results get counted. ChatGPT alone handles about 12% of Google’s search volume for queries people used to type into Google, according to Ahrefs traffic research, and the answers those users read are assembled from third-party sources. This guide covers what digital PR means now, why the target list changed, which campaign formats earn AI citations, and how to measure the work. The recommendations draw on our own studies of 153,425 and 42,971 AI citations.
What digital PR is now
Digital PR earns four assets at once: editorial links, unlinked brand mentions, third-party citations, and entity records such as Wikipedia pages and knowledge panels. Classic link building counts only the first. The other three now carry more weight, because AI engines select sources by brand presence rather than link equity. Across 75,000 brands, Ahrefs measured branded web mentions correlating at 0.66 to 0.71 with AI visibility. Domain Rating correlates at 0.266 to 0.326, and link metrics such as backlink count show very weak correlations across every AI system tested.
Digital PR for SEO historically meant link acquisition with a media relations wrapper. The same placements now also decide AI citations, which is a larger prize per placement. Traditional PR aims at audience reach and sentiment. Link-era campaigns aimed at referring domains and authority scores. A digital PR strategy for 2026 aims at the specific sources each AI engine pulls into answers for your category, and YouTube, Reddit, and Wikipedia dominate those sources alongside trade journalism.
| Link-metrics digital PR | Digital PR for AI search | |
|---|---|---|
| Primary target | High-DR domains in any niche | Outlets and platforms each engine cites in your category |
| Core asset | Guest posts and linkable pages | Data studies, expert quotes, product-source pages |
| Success metric | Referring domains, authority scores | Citation share inside AI answers, plus links |
| Mention without a link | Counted as a failure | Counted as a result |
| Refresh rhythm | Quarterly prospect lists | Monthly per-engine citation audits |
How a digital PR campaign runs now
The mechanics stay recognizable. The selection criteria at each step changed.
- Build the asset. A data study, an expert position with numbers behind it, or a source page journalists and engines can quote.
- Build the list. From observed AI citations in your category, per engine, plus the trade outlets those citations come from.
- Pitch. Exclusive angle to the outlets that already get cited, then broader outreach, then the press release to widen pickup.
- Seed the platforms. The subreddits, YouTube channels, and LinkedIn surfaces the engines cite for your category, through participation rather than placement.
- Measure citations. Re-run a fixed prompt set and count your mentions inside answers, next to the links and coverage you already report.
Why AI answers changed the target list
Our May 2026 study of 153,425 AI citations found that 76.95% of cited URLs are not in the organic top-10 for their query. Rankings and citations are separate contests. A campaign that only moves rankings leaves most of the citation surface untouched, which is why link building has been repriced rather than retired: links still handle crawl discovery and entity coherence, and they stopped being the score.
Each engine reads a different press list
ChatGPT cites sources in 96% of responses and averages five citations per response. Gemini cites in 82% and averages eight. Claude cites in 55% of responses and averages 13 when it does. The top cited domain differs too. Muck Rack’s report found ChatGPT leans on Wikipedia, Claude leans on PubMed Central, and Gemini leans on Reddit.
Journalism makes up 27% of cited sources, spread across more than 20,000 outlets. Axios appears in ChatGPT’s top three cited domains in 13 of 17 industries, the only journalism outlet to reach a top three on any platform. One placement in the right trade outlet now carries direct citation value on top of its link value.
Community platforms hold the other large share. Profound’s analysis of over 4 billion AI citations found Reddit is the single most cited domain aggregated across ChatGPT, Google AI Overviews, and Perplexity, at 3.11% of citations, ahead of YouTube at 2.13% and Wikipedia at 1.35%. YouTube mentions are the strongest single correlate of AI visibility at roughly 0.737 in the Ahrefs data, which is why video coverage belongs on a PR target list. These placements also keep paying out: Profound found the average Reddit post cited in 2025 is one year old, so a mention earned this quarter feeds answers well into next year.
The target list moves monthly
Engine source preferences shift fast enough to invalidate a quarterly plan. On 23 August 2026, ChatGPT cited reddit.com in 2 of 100 answers while Google AI Mode cited it in 50 of the same 100. The prompts were identical and the runs were an hour apart. A target list tuned to one engine does not transfer to another, and last quarter’s list may already be stale.
The new target list
Build the list from observed citations, per engine, per category. Ask each engine the questions your buyers ask and record which domains it cites. Six target groups recur across our tracking and the published studies.
- Trade press and business journalism. The 27% journalism share concentrates in category-specific outlets, with digital-native newsrooms like Axios overrepresented in ChatGPT.
- Wikipedia and entity records. ChatGPT’s most cited domain. AI engines use Wikipedia for entity grounding, so a clean page, correct Wikidata, and consistent naming pay off across every engine.
- Reddit and category subreddits. Engines treat niche subreddits as subject-matter sources for buying questions. The Reddit approach that earns AI citations runs on genuine participation, and its payoff varies sharply by engine.
- YouTube. The 0.737 correlation covers mentions in titles, transcripts, and descriptions. Creator reviews and podcast appearances are PR placements, and they compound across Google surfaces.
- LinkedIn. LinkedIn is one of the most cited sources for professional queries. Executive commentary and study summaries posted there get quoted directly.
- Review platforms. G2 and Trustpilot profiles feed AI brand citations on commercial queries, so profile completeness and review volume belong in PR scope.
The mix per engine matters more than the total. A brand strong on Reddit and weak in trade press will surface in Gemini and stay invisible in ChatGPT. Run the audit per engine before allocating budget.
Campaign formats that earn AI citations
Original data studies
Original research is the highest-ROI content format for AI citations: engines pull unique numbers and methodology-backed findings at far higher rates than opinion or summary content. The Princeton GEO paper, accepted to KDD 2024, measured content-side optimization boosting visibility in generative engine responses by up to 40%. A study also gives journalists a reason to cover you, which converts one asset into the earned coverage that drives the 84% share.
Make the study quotable at sentence level. Our sentence-level decode of 42,971 citations found cited sentences run 6 to 17 words, cluster in the top 35% of the page, and structured content gets cited 2.3x more often than prose. A study that earns citations meets a short spec:
- A number nobody else has, from data you own or collected yourself.
- Headline findings in the top third of the page, one fact per sentence.
- Methodology visible on the page, with sample sizes and dates.
- A chart or table outlets can republish with attribution.
- A named author a journalist can quote and an engine can attribute.
Expert commentary
Muck Rack’s query-type data shows industry trend questions drive journalism citations at more than double the rate of how-to questions. Commentary placements in trade coverage put your executives inside exactly those answers, and founder commentary earns AI citations in its own right. Reactive press requests are the cheapest entry: a two-paragraph response with a concrete number beats a canned statement.
Press releases that trigger coverage
A press release by itself produces almost no citations. The editorial coverage it triggers produces them at scale, which is the basis of the press release strategy for AI search. In the Muck Rack data, press releases appear almost exclusively in industry trend responses, at 3.5 times the rate they appear in best-of queries. Write the release to hand journalists a story: the dataset, the chart, the quote.
Product-source pages
Engines answer pricing and capability questions by quoting whichever page states the fact plainly. A product-source page gives them yours: current pricing, limits, integrations, one fact per sentence. Calculator and tool pages earn citations the same way, by producing numbers that exist nowhere else. Keep every fact you want quoted in visible page text.
How to measure digital PR now
Coverage reports counted links and estimated reach. AI answers add a citation measure, and it is the one that reflects the 84% earned-media share. Track five numbers.
- AI citation share. Your domain’s share of citations across a fixed prompt set, against named competitors. Share of voice in AI answers is the closest thing to a rank report this channel has.
- Per-engine mention rate. Score ChatGPT, Gemini, AI Mode, and Perplexity separately. The Reddit split above shows why an average across engines hides the story.
- Citation velocity. How fast new coverage starts appearing in AI answers is the leading indicator that a campaign landed.
- Earned mention count. Linked and unlinked, from standard media monitoring. Unlinked mentions still count toward the 0.66 to 0.71 mention correlation.
- Downstream lift. Branded search volume and AI referral sessions, reported as a trend rather than attributed per campaign.
Sample repeatedly before reading any of these numbers. Our variance study sent 28 identical questions to ChatGPT four times each: only 12% of credited sources appeared in all four runs, and 56% appeared in exactly one. Read pooled citation share across repeated runs, never a single spot check.
The tooling can stay simple. A spreadsheet, a fixed prompt set, and a monthly hour per engine cover a single brand. We run our own tracking with our open-source GEO/AEO tracker, which automates the prompt runs and the per-engine citation counts.
A 90-day digital PR playbook
This sequence fits a small team and produces its first measurable citation movement inside one quarter.
- Weeks 1 to 2: prompt audit. Ask each engine 50 buying questions for your category. Record every cited domain. That output is your target list, split by engine.
- Weeks 3 to 6: one data study. Build it from data you already hold: support tickets, usage logs, price tracking. Ship 10 to 15 quotable stats with the methodology on the page.
- Weeks 5 to 8: press outreach. Pitch the study to the trade outlets from your audit, exclusive angle first. Publish the release after the first placement to widen pickup.
- Weeks 7 to 10: entity cleanup. Fix your brand SERP and knowledge panel, Wikidata record, and naming consistency so new coverage resolves to one entity.
- Weeks 9 to 12: platform presence. Seed the study where your audit says it matters: the two or three subreddits engines already cite, one YouTube collaboration, a LinkedIn summary from a named executive.
- Ongoing: monthly re-audit. Re-run the prompt set, recompute citation share, and report it next to links and coverage. The target list will have moved; the August Reddit split proves how fast.
Digital PR used to sell influence that was hard to count. AI answers made the influence countable. Teams that run every campaign against a per-engine target list and report citation share will own the answers their buyers actually read.