Noma by Nurdd (NomaHQ)Noma
Sign up

Free playbook · PDF

Digital PR for AI search: how to earn the coverage ChatGPT and Gemini cite

A PR playbook for AI answers: build a media list weighted by what ChatGPT, Gemini, Perplexity and Claude cite, write pitches that carry brand, claim and category in one quotable sentence, follow disclosure rules, and measure coverage-to-citation.

  • Citation-weighted media list worksheet, scored at article level
  • Quotable-sentence builder and five pitch templates
  • Data-story planner with a publishable-method checklist
  • Coverage-to-citation tracker and a reporting format for the CCO
  • A 90-day PR plan, week by week

Get your free copy

Instant download, plus a copy by email. No spam: one email with your link, and a reply from a person if you write back.

The short answer

To get cited by ChatGPT and other assistants, earn coverage on the pages they already cite for your buyers' questions. Run those questions, record every cited article, and rank outlets by how often they are cited. Then pitch stories whose key sentence names your brand, the claim and the category together, and track which coverage reaches the answers.

For in-house PR and communications teams, and PR agencies

What you will learn

  1. Chapter 1

    Why third-party coverage shapes AI answers

    • Coverage reaches AI answers through training data, which is slow and invisible, and through search at answer time, which is fast and shows up as a citation.
    • Muck Rack's May 2026 study reported earned media at 84% of AI citations in its sample, and an academic study reported a strong bias towards earned media in AI search; both measure averages, not your category.
    • Measure citations, mentions and whether the answer carries your claim, because only the last one changes a buyer's shortlist.
    • A single precise sentence that names your brand, claim and category is what an assistant can lift; general praise gives it little.
    • Check whether target outlets allow the crawlers of the engines your buyers use, since a blocked crawler means no citation on that engine.

  2. Chapter 2

    Build a citation-weighted media list

    • Build the media list from the questions buyers ask assistants, then record what the engines cite, rather than starting from outlets you already pitch.
    • Run each question several times on each engine and record full article URLs, because cited sources vary from run to run.
    • Score articles by priority-weighted citations, then roll up to outlets and authors.
    • Sort targets into Defend, Win, Fix and Watch so the score leads straight to an action.
    • Refresh the list each quarter, and decide who owns review platforms and comparison sites that are cited but are not traditional press.

  3. Chapter 3

    Pitch construction: one quotable sentence

    • Write the sentence you want repeated first: brand, category in buyers' words, one specific claim, and its evidence or conditions.
    • Keep the brand name in the same sentence as the claim so the two are not separated when the sentence is lifted.
    • Place the sentence in the pitch summary, the spokesperson quote, the boilerplate and your own newsroom, in consistent words.
    • Back every claim with evidence a journalist can verify quickly, and drop superlatives you cannot prove.
    • Press releases are your own statement: use them as the record and avoid keyword-anchored links, then pitch for independent coverage.

  4. Chapter 4

    Stories engines can cite: data, commentary and analysts

    • Original data earns coverage in which your brand and a specific finding travel together, and can be cited long after the news peg.
    • Publish a methods page with sample, dates, questions, funder and limits before pitching any data story.
    • Named spokespeople with defined topics and prepared positions keep you in the recent journalism engines cite for trend questions.
    • If analyst pages appear in your citations, brief analysts with the same category words and claim you use with the press.
    • Score story ideas on question fit, outlet fit, claim fit, evidence and shelf life before investing in them.

  5. Chapter 5

    Podcasts, video and Wikipedia

    • Podcast and video appearances reach text-based systems through transcripts and captions, so check and correct them after every appearance.
    • YouTube and Apple Podcasts both let creators supply their own captions or transcripts in place of automatic ones.
    • Say your quotable sentence once, in full and with the brand name, during any recorded appearance.
    • On Wikipedia, anyone connected to the company must disclose and propose changes on the talk page with {{edit COI}} instead of editing directly.
    • Wikipedia's notability guideline excludes press releases and sponsored content, so independent earned coverage is what an editor can cite.

  6. Chapter 6

    Ethics and disclosure

    • Paid placements must be labelled: the FTC's native advertising guidance and the UK CAP Code rules 2.1 and 2.4 both require it.
    • The FTC's Endorsement Guides, revised in 2023, require clear disclosure of material connections such as payment, free products or employment.
    • Google's spam policies treat paid links with optimised anchor text, including in press releases, as link spam, and describe site reputation abuse.
    • Never fabricate data, quotes or experts, never post without disclosure, and never promise that a placement will produce an AI citation.
    • Agencies share responsibility for disclosure, so write the rules into the statement of work.

  7. Chapter 7

    Measure coverage-to-citation, and a 90-day plan

    • Track every piece of coverage through published, fetchable, cited, mentioned and claim carried, because each failure has a different fix.
    • Report pickup rate, mention rate and claim carry-through as fractions as well as percentages, per engine.
    • Compare targeted questions with untargeted ones before claiming a campaign caused a change.
    • A one-page report to the CCO covers questions and engines tracked, the measures, direct evidence of pickup, and next steps.
    • Run a 90-day cycle: baseline and media list first, corrections and stories next, then rerun and refresh.

6 worksheets inside

The chapters explain the method. The worksheets are what you fill in to run it: print them, or copy the tables into a document of your own.

  • W1Citation-weighted media list
  • W2Quotable sentence builder and pitch templates
  • W3Data story planner
  • W4Ethics and disclosure checklist
  • W5Coverage-to-citation tracker
  • W690-day PR plan and next step
Get the playbook

Questions

How do I get my brand cited by ChatGPT?
ChatGPT search cites pages it finds through search, so earn coverage on the pages it already cites for your buyers' questions. Run those questions, record the cited articles, and pitch the writers and outlets that appear most often. Make sure the coverage names your brand, category and claim in one sentence, and check the outlet does not block OAI-SearchBot, which OpenAI says is required to appear in ChatGPT search answers.
Which publications does ChatGPT cite?
It depends on the question and the category, and it changes over time, so the only reliable answer is to run your own buyers' questions and record the citations. As one published data point, Muck Rack reported in May 2026 that Axios appeared in ChatGPT's top three cited domains in 13 of the 17 industries it studied. Your list will differ, and specialist trade titles and review sites may matter more than national outlets.
Does earned media really affect AI answers?
Published research suggests it is a large part of what assistants cite. Muck Rack's May 2026 study reported earned media at 84% of citations in more than 25 million links from ChatGPT, Claude and Gemini, and an arXiv study from September 2025 reported a strong bias towards earned media in AI search compared with Google. Both are averages across their samples, so measure your own category.
What is digital PR for AI search, or PR for GEO?
It is PR planned and measured against AI answers. The media list is built from the articles assistants cite for buyer questions, pitches are written so a key sentence carries the brand, category and claim together, and success is measured by whether coverage is cited and whether answers repeat the claim, per engine.
Do press releases help with AI visibility?
They are public pages that engines can read, and Muck Rack's May 2026 study found them cited almost exclusively in answers to industry-trend questions. Because a release is your own statement, it is a weaker source than independent coverage for questions about which product to choose. Use releases as the record of news, avoid keyword-anchored links, and pitch for independent coverage.
Can we pay for articles to appear in AI answers?
You can buy sponsored content, but it must be clearly labelled as advertising under FTC guidance in the US and the CAP Code in the UK. Muck Rack reported paid and advertorial content at 0.3% of AI citations in its May 2026 sample. Paid links for ranking also break Google's spam policies unless they are marked as sponsored or nofollow.
Can our PR agency edit our Wikipedia page?
Wikipedia's conflict of interest guideline strongly discourages connected editors from editing directly, and its paid-contribution policy requires agencies to disclose both themselves and the client. The agency should disclose, then propose specific changes on the talk page with the {{edit COI}} template, citing independent coverage rather than press releases.
How do I measure whether coverage reached AI answers?
Track each piece of coverage through five stages: published, fetchable, cited, brand mentioned and claim carried. Report pickup rate, mention rate and claim carry-through per engine, and compare questions your campaign targeted with ones it did not, so you do not credit the campaign for changes it did not cause.
Is Earned Media for AI Answers free?
Yes. Give a first name and a work email and the download starts at once. We also email you the link, and one person from the Noma team follows up a day later; reply and say so if you would rather not hear from us.
Why do you ask for a work email?
The playbooks are written for teams at companies, and a work address tells us which company is reading, so a follow-up can be about your brand rather than generic. Personal and temporary inboxes are not accepted.
What format is it?
A PDF of 40 A4 pages, laid out to print, with worksheets you can fill in by hand or copy into a document of your own.

More playbooks

Updated 20 October 2026 · Written by the Noma team · All resources