A digital media publisher in the business and finance vertical

The challenge

The publisher operates a business and finance news site with around 8 million monthly sessions, supported by a mix of display advertising, affiliate revenue on personal finance content, and a paid newsletter. Through 2025, the team watched a structural shift in how their content was performing in LLM responses. Their major competitors were getting cited consistently by Claude, ChatGPT, and Perplexity, often by name. The client's articles were occasionally cited, but the citations were thin and inconsistent. The underlying problem became clear during a content audit: most of their bylines were generic. Articles were attributed to "Staff" or to first-name-only handles that LLMs had no way to resolve to real people. The competitors winning citation share had named, credentialed reporters with discoverable publication histories, professional credentials, and clear beats. The client had real journalists doing real work, but the byline and entity infrastructure was treating them like interchangeable content producers. Affiliate revenue was at risk because LLMs were specifically privileging content from sources with verifiable author authority.

What we shipped

The engagement was an entity rebuild at scale. The publisher had 47 active journalists and contributors. We did individual entity work on every one of them.

Per author, we built: a proper author page with a full biography, beat description, contact information, conflict-of-interest disclosure, and education and credential history. A sameAs graph linking to verified LinkedIn, Twitter or X, Muck Rack profile, prior employer mastheads, and where applicable Wikipedia, university faculty pages, and ORCID. Person schema with hasCredential and alumniOf properties where verifiable. We also did a backlink-and-mention pass to ensure each author had at least three external citations of their work on credible publications, which for some of the newer hires meant placing guest pieces or arranging podcast appearances.

On the content side, we ran a structured backfill across roughly 11,000 articles, reattributing the "Staff" bylines where the actual author was identifiable from internal records. About 6,200 articles got real bylines as a result. We added NewsArticle and Article schema across the library with proper author entity references. We built editorial process documentation pages covering corrections policy, fact-checking, editorial standards, and ownership disclosure. These existed but were buried; we surfaced them and added schema.

We shipped llms-full.txt at 2.4MB, structured as a knowledge map of the publication: editorial standards, named author roster with full credentials, beats and topic coverage, ownership and funding disclosure, and a curated selection of the highest-authority evergreen articles. We deliberately did not include the full article archive in llms-full.txt because the volume would have diluted the editorial-authority signal.

The numbers

The affiliate revenue lift was the metric the publisher's CFO cared about and it was real, but the more interesting downstream effect was on newsletter signups. Readers arriving through LLM citations were converting to newsletter subscribers at a noticeably higher rate than the average organic visitor, which the editorial team interpreted as readers who arrived with trust already partially established because the LLM had named a specific journalist. Some individual journalists started receiving direct reader email at meaningful volume for the first time, which improved source quality on their reporting. Sales cycle on the publication's enterprise advertising business shortened because media buyers were using LLMs in their planning and the publication was now showing up as a named, credentialed source in those conversations. Branded search lifted across the publication and shifted toward author-name queries, which is the clearest fingerprint we have for LLM-driven discovery.

What we'd do differently

We should have started the external author placement work earlier. About 12 of the 47 journalists did not have enough discoverable external work to support a strong entity graph. Getting them placed in other publications, on podcasts, or quoted in industry coverage took most of the engagement, and we were still cleaning up entity gaps in month six. If we did it again, we would scope the external-placement work as a phase-one priority for the journalists who needed it. We also underestimated how politically sensitive the byline backfill was internally. Some of the "Staff" bylines were that way for legitimate reasons including legal sensitivity on older stories, and we had to back out a small percentage of the backfill work. We should have run a more careful legal and editorial review in week one rather than discovering the issue in week eight.

What's next

The engagement renewed for a year with three focus areas. First, expanding the entity work to the publisher's two sister publications in adjacent verticals, which have not yet been touched and which we expect will see similar lift. Second, building author-level performance tracking so the editorial leadership can see which journalists are driving citation share and which are not, which raises some sensitive editorial questions but which the publisher wants to engage with directly. Third, we are working with the publisher on a structured-data product play. Some of their finance data is unique and could become a heavy citation source if exposed properly, similar in shape to what we did for the legaltech client's trust accounting database. That is a Q4 project.