PR and SEO work together because both depend on credible third-party pages that mention a brand. Muck Rack's May 2026 study of more than 25 million links found that earned media made up 84% of AI citations across ChatGPT, Claude and Gemini, and paid or advertorial content made up 0.3%. A joint plan starts with one shared question: which outlets mention the brand, and do those pages get read, ranked and cited?
This guide shows a communications lead or an SEO lead how to score any placement for both search and AI value, split the work between the two teams, and report one shared number. It draws on public research from Muck Rack's What Is AI Reading? series, Ahrefs, Pew Research Center and Google's own documentation, and it marks which findings are correlations rather than proven causes. Readers who want the search basics first can start with the search engine optimization guide.
What do PR and SEO each produce, and where do they overlap?
PR produces third-party coverage, SEO produces findable pages, and the overlap is the set of outside pages that mention or link to a brand. Those outside pages are the raw material for both disciplines, which is why the two teams often chase the same outlets without knowing it.
| Discipline | Primary output | Usual measure | Shared output |
|---|
| PR | Coverage, quotes and mentions in outlets | Placements, reach, sentiment | Third-party pages that name the brand |
| SEO | Indexable pages, internal links, technical fixes | Impressions, clicks, rankings | Pages worth linking to and quoting |
| AI visibility (GEO) | Pages and mentions that AI answers cite | Share of AI answers that name the brand | The same third-party pages, read by AI systems |
Earned media is coverage that a publication or other third party chooses to publish without payment. SEO, short for search engine optimization, is the work of making pages findable in search. GEO, short for generative engine optimization, applies similar work to answers written by tools such as ChatGPT, Claude, Gemini and Google AI Mode. An AI citation is a source link that an AI answer shows next to its text. For a fuller definition of the newest term, see what generative engine optimization is, and for the older discipline, see what earned media is.
One definition matters before any number below. Muck Rack counts journalism, academic research, government sources, encyclopedic sites and third-party corporate content as earned media. Its 84% therefore does not mean 84% of citations come from news stories. In the December 2025 edition, Muck Rack reported that journalism specifically made up 20% to 30% of citations, according to its Generative Pulse Summit recap.
The best public data says that unpaid third-party pages supply almost all of the sources AI answers cite, while paid placements supply almost none. Muck Rack's May 2026 edition of What Is AI Reading? reports 84% earned media and 0.3% paid or advertorial across more than 25 million links from ChatGPT, Claude and Gemini. The two earlier editions, in July 2025 and December 2025, showed the same pattern at 95% and 94% non-paid. EPR's own roundup of these studies is in earned media and AI citations.
Two further findings from the December 2025 edition change how a PR team should work. Muck Rack found only a 2% overlap between the journalists brands pitch most and the journalists AI systems cite most. It also reported that press release citations rose fivefold between July and December 2025, with releases that carry more statistics, bullet points, action verbs and objective sentences cited at higher rates.
Muck Rack sells Generative Pulse, a product that tracks this exact behavior, so its reports are vendor research. The methodology is published, and the report notes that the behaviors it observed may shift as models are updated or retrained. Read the percentages as a snapshot of three AI tools in 2025 and 2026, not as a fixed rule.
Ahrefs' study of 75,000 brands points the same way from the SEO side. In its AI brand visibility correlation study, branded web mentions correlated with AI visibility at 0.66 to 0.71 across ChatGPT, Google AI Mode and Google AI Overviews, and YouTube mentions correlated at about 0.737, the highest measured. These are correlations. Well-known brands tend to have more mentions, more searches and more AI visibility all at once, so the data cannot separate the effect of mentions from the effect of brand size.
Clicks fall even when a brand is cited. Pew Research Center analyzed 68,879 Google searches by 900 U.S. adults in March 2025 and found that users clicked a traditional result link in 8% of visits that showed an AI summary, compared with 15% of visits that did not. A Pew Research Center short read gives the full method. A joint program that counts only clicks will undercount the value of coverage that shapes the answer itself.
Do press links still help Google rankings?
Links from press coverage can support rankings when a publication earns them by choice, and Google's spam policies define the versions that cross the line. Google's spam policies classify links created primarily to manipulate rankings as link spam. The listed examples include buying or selling links for ranking purposes and excessive link exchanges.
The same policy covers two habits common in PR work. It names advertorials or native advertising where payment is received for articles with links that pass ranking credit, and it names links with optimized anchor text in articles, guest posts or press releases distributed on other sites. Google says paid and sponsored links are acceptable when they carry a rel="nofollow" or rel="sponsored" attribute.
The practical rule for a pitch is to ask for accurate attribution and a link to the page that holds the evidence, and never to specify keyword anchor text. A journalist who cites a research page links because the page supports the story. That is an editorial link, and it is the kind a brand can earn repeatedly by publishing data worth citing. The Links report in Google Search Console lists which sites link to each page, which shows whether a campaign produced links or only mentions.
How do you score a placement for search and AI value?
A placement scorecard rates each piece of coverage on six questions so that a trade-press quote and a national feature can be compared on one scale. EPR uses the scorecard below as an editorial tool. It is a planning aid built from the research above, not a published ranking formula from Google or any AI company.
| Question | Strong placement | Weak placement |
|---|
| Which outlet published it? | A named newsroom with an editor and a topic beat | An unknown site with no staff or editorial standards |
| What kind of link does it carry? | An editorial link to a page with the evidence, or a clear unlinked mention | A paid link without a sponsored attribute, or optimized anchor text |
| How prominent is the brand? | The brand or its data is the subject of the piece | A one-line mention in a list |
| Who wrote it? | A reporter who regularly covers the topic | A contributor with no link to the topic |
| Does the topic match buyer questions? | The page answers a question buyers ask | The page covers an unrelated subject |
| How recent is it? | Published within the last year | Several years old and not updated |
Illustrative scenario: a software company earns two placements in the same quarter. The first is a national business reporter quoting its chief technology officer on a security trend, with no link. The second is a contributor post on a general-interest site, bought through an agency, with a keyword link and no sponsored attribute. The scorecard rates the first as strong on outlet, prominence, author and topic, and the second as weak on all four plus a link that Google's policy treats as a risk. The first placement is the one worth repeating, even though it brought no link.
How should PR and SEO teams split the work?
PR owns the pitch and the journalist relationship, SEO owns the destination page and the measurement, and both own the topic list. The split matters because coverage helps most when a journalist has something specific to cite, and that something is a page the SEO team keeps accurate and crawlable.
| Step | PR team | SEO team | Shared output |
|---|
| 1. Pick topics | Story angles that reporters cover | Queries buyers search and questions they ask | One topic list with demand and news value |
| 2. Build the source page | Supplies data, quotes and a named spokesperson | Publishes an indexable page with the data and clear headings | A page worth linking to |
| 3. Pitch | Targets reporters who cover the topic | Supplies the URL and supporting facts | A pitch list organized by beat |
| 4. Release | Writes with statistics and objective sentences | Makes the release page indexable and avoids keyword anchors | A release that fits Google's link policy |
| 5. Report | Keeps the coverage log and quality scores | Tracks Search Console and AI answers | One shared report |
Step three is where the Muck Rack finding on journalist overlap applies. If only 2% of the most-pitched journalists are the ones AI systems cite most, then the SEO team's citation data is a pitch list the PR team does not yet have. Running a set of buyer prompts and recording which outlets the answers cite shows which reporters and publications to approach next. Guides on how to get press coverage and link building and digital PR cover the pitching mechanics, and the digital PR tools page lists software for tracking results. Teams choosing a media database can compare options in the digital PR databases comparison.
Step four follows the same logic. Muck Rack's data shows that structured releases get cited more often, and Google's policy shows that distributed releases with optimized anchors count as link spam. A release that states its facts in plain sentences and links to the source page with a descriptive, non-keyword anchor serves both findings. EPR's press release retrieval study reports that 47 of 250 wire-distributed releases were cited by AI engines within 90 days. The writing rules are in how to write a press release, and the buying guide is press release distribution services.
How do you build a pitch list from cited sources?
A pitch list built from cited sources starts with the outlets that already appear in AI answers on the brand's topics, then narrows to the reporters at those outlets who cover the subject. The method needs the prompt set from the measurement step below.
- Run each buyer prompt on ChatGPT, Gemini, Claude and Google AI Mode, and copy every cited domain into a spreadsheet.
- Count how many prompts each domain appeared in, and sort the list from most to least frequent.
- Open the cited pages and note the byline. A domain that appears often because of one reporter is a reporter target, and a domain that appears because of a reference page is a different kind of target.
- Remove domains the brand cannot realistically earn coverage in, such as encyclopedias or government sites, and keep outlets with named editorial staff.
- Pair each remaining reporter with the source page that fits their topic, and write the pitch around that page's data.
The result differs from a standard media list in one respect. A standard list is built from who the team knows or who covers the industry in general. This list is built from where answers already come from, which is the gap the Muck Rack overlap figure points to. Muck Rack's July 2025 report also found that query wording changes which sources get cited, so run each prompt in two or three phrasings and keep domains that appear across them.
Review the list every quarter. Outlets enter and leave answers as models update and as new coverage appears, so a list that was accurate in January may miss outlets by June. A quarterly rerun keeps the pitch plan tied to what the tools currently read.
What should a pitch contain to earn a citation?
A pitch earns a citation when it gives the reporter a specific, checkable fact and a clearly identified source to attribute it to. Four habits make that more likely, and each one is a practice EPR recommends, not a finding from the studies above.
| Habit | What to do | Why it helps |
|---|
| Lead with a number | Build the pitch around one statistic from the brand's own research, and link the page that holds it | A reporter can write the story around the figure, and a reader or an AI answer can trace it to the source page |
| Name the spokesperson the same way every time | Use one name, one title and one affiliation across every placement | Systems that match names can read inconsistent versions as different people |
| Ask for context in the piece | Provide a two or three sentence bio the reporter can use next to the quote | A quote with its credentials attached can be attributed to a real person and organization |
| Keep a placement page | Publish a page on the brand's site that lists major coverage with links and dates | It gives readers and crawlers one place to verify who the spokesperson is and where the work appeared |
Two older habits work against this. Sending one identical email to twenty reporters tends to produce coverage in outlets that never show up in the prompt data, which the cited-source list above is built to avoid. Announcing news without a verifiable claim leaves nothing for a reporter or an answer to attribute.
What does bad PR for SEO look like?
Bad PR for SEO is coverage bought or built mainly to create links or mentions, and Google's spam policy names several versions of it. Paid links without a sponsored attribute, link exchanges between partner sites and keyword-rich anchors in guest posts are the common forms. Sites that violate the policy may rank lower or stop appearing in results, according to Google.
Paying for coverage also appears to be a weak route into AI answers. Muck Rack found paid and advertorial content at 0.3% of citations in May 2026, so a brand that buys placements in the hope of appearing in ChatGPT or Gemini is spending where the sampled citations rarely come from. The finding describes citation patterns in that sample and does not prove that paid content could never be cited.
A quieter failure is pitching the wrong people. The 2% overlap figure means most pitch lists are built from habit, not from evidence about which outlets feed answers on the brand's topics. A team that rebuilds its list from cited sources, and drops outlets that never appear, spends the same effort on more useful targets.
How do you measure PR and SEO as one program?
A joint program needs one shared measure: the share of answers, across a fixed set of buyer prompts, that name the brand and cite a page the brand has earned. Choose 25 to 50 prompts that real buyers ask, run them on a set schedule across ChatGPT, Gemini, Claude and Google AI Mode, and record whether the brand appears and which domains the answer cites.
Run each prompt more than once and record the dates. Muck Rack's report states that its observed behaviors may shift as models are updated or retrained, which means a single run is a sample and a month-to-month change can reflect a model update instead of a campaign. Pair the prompt data with Search Console so that a drop in clicks is read against the Pew finding that AI summaries lower click rates.
To calculate the share, divide the number of prompt runs in which the answer names the brand by the total number of runs, and report it with the number of runs behind it. A result of 12 named answers out of 60 runs is a 20% share, and the 60 matters because a share built on six runs can move by 17 points on one answer. Keep a second figure for runs that cite a page the brand has earned, since being named without a citation and being cited are different outcomes.
| Measure | Where it comes from | What it shows |
|---|
| Branded query impressions and clicks | Google Search Console | Whether coverage raises demand for the brand |
| Sites linking to the source page | Search Console Links report | Which outlets linked, and which only mentioned |
| Share of prompt answers naming the brand | Repeated prompt runs or a monitoring tool | Whether AI answers include the brand |
| Domains cited for the prompt set | The same prompt runs | Which outlets to pitch next |
| Placement quality score | The scorecard above | Whether coverage is the kind that gets cited |
Earned media measurement has a longer history than AI citation tracking, and the prompt-set share should sit inside it instead of replacing it. The measurement hierarchy used by AMEC's Barcelona Principles separates outputs (what the brand produced), outtakes (what audiences received), outcomes (what audiences thought or did differently) and impact (the business result). Placement counts and quality scores are outputs, prompt-set share is closer to an outtake, and pipeline or revenue attribution is impact. A report that spans several of those levels is easier to defend to a finance lead than one that counts placements alone.
Executives often ask for one slide. The prompt-set share is the number to put on it, with the placement quality average beside it. The briefing format in how to explain SEO, GEO and AI visibility to executives shows how to present both.
What should a team do in the first 90 days?
A team can start with three 30-day phases that move from audit to pitching to measurement. The plan below needs no new tools beyond Search Console and a spreadsheet, and it produces a baseline that later campaigns can be compared against.
| Days | Actions | Output |
|---|
| 1 to 30 | Build the prompt set. Run it and record which brands and domains appear. Score the last 12 months of coverage with the scorecard. List pages that journalists could cite. | A baseline prompt-set share and a coverage score |
| 31 to 60 | Publish or update two source pages with original data. Rebuild the pitch list from cited outlets. Pitch reporters with the page URL and a named spokesperson. | Source pages, a new pitch list and a pitch log |
| 61 to 90 | Re-run the prompt set. Check Search Console for new linking sites and branded queries. Drop outlets that produced nothing. | A before-and-after report and a revised target list |
The order matters. Auditing first shows which outlets already shape answers on the brand's topics, so pitching begins with evidence. Publishing source pages before pitching gives reporters a specific, accurate page to cite instead of a homepage. Re-measuring last shows which of those moves changed the answers, and which were activity without effect.
