Anyone who writes a text without a byline asks the reader to take their word for it. A human sometimes does that. An AI system that compiles an answer and shows sources alongside it almost never does: it attaches its own credibility to the sources it chooses.

That's why the question of who says something now weighs just as heavily as what is said. The measure for this has a name.

01 / the four lettersWhere E-E-A-T comes from

E-E-A-T comes from the guidelines Google uses to assess the quality of pages. The letters stand for Experience (have you been through it yourself), Expertise (do you know what you're talking about), Authoritativeness (do others recognise that) and Trust (is what's written accurate, and can it be checked who says it). According to Google's own guidelines, the last one weighs most heavily: without trust, the other three matter little. In one overview, with what each letter means on an ordinary business site:

For a long time this was mainly a concern for sites about health and money. That has shifted. Google states that its AI features rest on the same ranking and quality systems as ordinary search results, and ChatGPT, Claude and Perplexity also choose their sources on signals of trustworthiness. It's plausible that a language model citing a source takes on a risk to its own name by doing so, and therefore behaves more cautiously than a results list ever did.

02 / in practiceWhat it looks like on an ordinary business site

Four things, and none of them require a technical project.

A named author. Every content page carries a name, with one line explaining why that person is qualified to write about this. Not a life story, but an identifiable background.

Genuine experience in the text. The difference between "many businesses struggle with this problem" and a description of what you saw with clients yourself. Anyone can write the first, which is why it's worthless as a signal. Only someone who was there can write the second.

Sources that can be checked. For a machine, a claim with a traceable source and a year is a different object than the same claim without one. How to build that into your texts is covered in how you write content AI can cite.

Visible dates. A publication date and an identifiable moment of last maintenance. Outdated information that's left standing drags down trust in the rest of the site.

A reader forgives an anonymous text. A language model skips it.

03 / the figureWhat a named expert delivers

This effect has been measured. The GEO research from Princeton (KDD 2024) measured an effect from adding citations from relevant sources, such as a quoted statement from an expert with a name and title in the text itself.

The reasoning behind it is straightforward. An AI system that has to choose between two comparable passages picks the one with an identifiable sender attached, because that choice is easier to justify. It works the same way as a journalist who would rather cite a named source than an anonymous one.

For a small business this is more of an advantage than a burden. A large organisation writes from behind a brand name; an entrepreneur can attach every text to a real person with real experience. That signal already exists, it just needs to appear on the page. How this fits into the bigger picture is covered in how you get mentioned in ChatGPT answers.

04 / the limitWhen this isn't your first priority

Three situations in which you're better off letting E-E-A-T rest for now.

Your site is technically unfindable. If search engines and AI crawlers can't read your pages, an author block won't change that. Access first, trust second.

There's no content yet to put a name to. An author byline on three thin pages adds nothing. E-E-A-T strengthens content that already exists; it doesn't replace content.

Your customers choose based on something else. For some local services, a review from a neighbour carries more trust than any author page. In that case your business profile with reviews deserves attention before your blog does.

If your site doesn't fit any of the three, the order is simple: put names to what's there, add sources and dates, and from now on write from what you've experienced yourself.

This text was produced with AI assistance and checked and approved by a human before publication.