For many buyer questions, AI engines answer from what others say about a brand rather than from what the brand says about itself. Earned authority is the work of being present, and described accurately, in those other sources.
PR teams already know how to earn coverage. What changes is how targets are chosen and how results are judged: both now come from the tracked answers.
When is earned work the right lever?
Chapter 5 sorts every question by who its answers cite. Earned work is the lever for one group: the questions whose answers cite third-party sites, such as publications, review platforms, forums, and directories.
Earned media is coverage by independent publishers, creators, and communities that the brand neither paid for nor owns. On these questions engines are looking for third-party validation, independent confirmation of what a brand claims, and another page on the brand's own site cannot supply it.
The questions also have to be inside the brand's market, under the scoping rule from chapter 3. Coverage that wins answers to questions the brand cannot serve does not reach a buyer.
Which sources belong on the target list?
The target list comes from the answers, not from a ranking of prestigious outlets. For each question in the third-party group, the cited domains show which sources the engines already trust on that subject.
Three readings of those domains shape the list:
- Sources that mention competitors and not the brand. These are the clearest gaps, because the engine already uses the source and the brand is missing from it.
- Sources cited across many questions. A source that shapes answers on several questions is worth more effort than one cited once.
- The kind of source. A publication, a review platform, and a community forum each call for different work, from pitching a story to asking customers for reviews.
One question's row on the list might read like this, with source types standing in for real names:
| Engine | Sources cited for the question | Brand present |
|---|---|---|
| Engine A | A trade publication, a review platform | No |
| Engine B | A community forum, the same review platform | No |
| Engine C | A competitor's own guide, an industry directory | No |
The review platform appears on two engines and omits the brand, so it heads the list. The forum matters only for one engine, and the competitor's guide is an owned-content gap that belongs back in chapter 5.
Digital PR is the work of earning coverage and links in those sources. The list tells the PR team where to aim; how to earn the coverage remains their craft.
Why does the list differ by engine?
Engines do not draw on the same sources. Asked the same question every day, Google's own three surfaces shared only a handful of the domains they cited, as Gemini, AI Overviews, and AI Mode cite differently shows.
A single blended list hides that. A target list read per engine shows which sources matter where, and lets a team put effort first into the engines its buyers use most.
How is the target list kept current?
Cited sources rotate. Many domains appear in a single answer and never again, so a source cited once is weak evidence and a source cited across weeks is strong evidence.
The list is therefore re-read every measurement period, from the same prompt set, rather than built once. Sources that keep appearing stay at the top, and sources that fade drop down before effort is spent on them.
Coverage that lands is checked the same way: whether the source still appears in the answers after the placement, and whether the brand now appears with it.
What does the knowledge graph add?
A knowledge graph is a database of entities and the relationships between them, such as a company, its products, and its founders. Reference sources that feed knowledge graphs give engines a stable account of basic facts about a brand.
That stability matters most for correcting errors. Chapter 1 explains that a brand hallucination with no cited source comes from the model's memory, and changes only as later models learn from a web that states the facts consistently. Accurate, consistent facts in reference sources are part of that correction.
Which signals should earned work move?
Earned work shows in the measurement differently from owned content, and judging it by the wrong signal makes it look like a failure.
| Signal | What earned work should do to it |
|---|---|
| Mention rate | Rise on the targeted questions, as engines pick up the brand from the new coverage |
| Share of voice | Rise against the competitors who held those sources |
| Cited sources | Show more third-party pages that mention the brand |
| Citation share of the brand's own pages | Possibly no change, because the citation goes to the publication |
The last row is the one that misleads. A successful placement is often cited as the publication's page, not the brand's, so the gain shows in mentions and share of voice. The same rules for judging change apply as for content: per engine, against the baseline, and larger than the usual movement.
What does earned authority not decide?
Earned authority decides how others describe a brand to the engines. Turning the measurement, the access work, the content, and the earned work into one repeating programme is the subject of the next chapter.
Terms in this chapter
- Earned media
- Coverage of a brand by independent publishers, creators, and communities that the brand neither paid for nor owns.
- Third-party validation
- Independent confirmation of a brand's claims from sources the brand does not control, such as reviews, press, and analyst coverage.
- Digital PR
- Earning coverage and links from independent publications to build a brand's presence in the sources AI engines read.
- Knowledge graph
- A database of entities and the relationships between them.
Zumi shows which third-party sources AI engines cite for a category and which of them mention competitors but not the brand, so earned-media effort goes where engines look. Book a demo