How EAT 2.0 Builds Authority That AI Cannot Flatten

What the reader detects under the surface

A buyer who has been reading well-written content for two decades can tell, inside the first three paragraphs, whether a real person was on the other side of the page or whether the page was generated to look like a person was. The tell varies by reader. The recognition is universal.

In 2026, that recognition matters more than the four signals Google has spent the last seven years teaching its raters to score. Surface credentials. Structured authority. Citation networks. Trust markers. A modern AI model passes all four at near-zero cost. The reader does not. The reader detects the absence under the surface even when they cannot name what is missing.

What they are detecting is the human layer Google’s original EAT framework was never asked to measure. Three components, none of them fakeable at scale: empathy, authenticity, transparency. EAT 2.0 is the operator’s response to a buyer who can now tell.

How EAT got here

In 2018, Google rolled out the Medic Update. The update penalized health and medical sites whose content could not be tied to qualified expertise, after Google had been watching too many pages publish health advice no qualified clinician would have signed off on. After Medic, the engine stopped pretending the surface of a page could be evaluated independently of who wrote it.

Out of Medic came EAT. Expertise, Authoritativeness, Trustworthiness. Three quality signals named in Google’s Search Quality Rater Guidelines, the published rubric Google’s human raters use to spot-check whether the algorithms are surfacing the right kinds of results. EAT is not a direct ranking factor. The algorithm does not measure it line by line. The algorithm learns from rater evaluations and surfaces results that match what the raters scored highly. The practical effect on visibility is the same.

Then EAT became E-E-A-T. Google added a second E for Experience, because some of the most helpful content on the web was being written by people who had lived a situation without holding a credential for it. A cancer survivor writing about treatment side effects. A parent writing about a specific developmental disorder. The lived experience was its own kind of authority, and the four-letter version of the framework named it explicitly.

After Medic, the doctors started calling. Plastic surgeons, dentists, specialists across the medical world wanted help with their EAT signals, and that work became a meaningful slice of my consulting practice for a few years. The mechanics of how I rebuilt their credibility surfaces sit in the FORCEPS framework. What matters here is what the doctors signaled. Authority had become a layer Google’s raters were grading on, and the algorithm followed. Operators who took it seriously earned the citations and the clicks. Operators who did not lost them.

The problem with EAT 1.0 in 2026 is not that the four signals are wrong. They are still what the raters score and what the algorithm follows. The problem is that AI now produces content that passes the EAT 1.0 surface test at near-zero cost. The credentials look right. The references look right. The structure looks right. The bibliography looks right. The reader still feels the absence. EAT 2.0 names what is missing.

Empathy

Empathy on the page is the reader catching the operator’s prior recognition of a situation the reader is currently inside. Not a “we understand” sentence. The recognition that makes the reader stop reading for half a beat and say, this person has sat where I am sitting.

This is the surface of the move the QUEST formula names as Understand. The reader who feels read stays. The reader who feels misread leaves, and the leaving is permanent in that moment, because nothing the page says after the misread will reach that reader again.

You cannot fake empathy at scale. It either lives in the work or it does not, and the binary is the part the model cannot manufacture. An operator who has sat across from the buyer carries the language in their tissue. An operator who has not, has nothing to imitate. The model can mimic the surface of empathy. The recognition empathy is built on has to come from somewhere outside the model’s training corpus, which is to say, from someone who was in the room.

The reader who feels recognized stays for reasons unrelated to the conscious decision to keep reading. Recognition lowers the resistance to the rest of the page, because the page has told the reader, accurately, who is on the other side. The work that follows gets evaluated on whether it earns the recognition, rather than on whether it deserves the attention. Attention has already been granted. The work decides what to do with it.

Authenticity

Authenticity is showing up on the page as a recognizably real person rather than as a brand-shaped surface.

Jessica Jensen, the CMO of LinkedIn, said it on the Uncensored CMO podcast. The posts performing best on the platform read as human, personal, sometimes vulnerable, sometimes whimsical. The platform’s own data points at what the framework points at. Surfaces written as a person outperform surfaces written as a brand. The reader can tell.

My own LinkedIn is the authenticity practice live. I write about powerlifting. I write about drumming for Nelson Colt, the country band I sit behind the kit for. I wrote about a recent emergency surgery for a bowel obstruction and turned the experience into business lessons about diagnosis, risk, and the things that get ignored until they cannot be ignored. None of those posts began as marketing. All of them did marketing’s work, because the surface was unmistakably mine.

The fractional buyer is reading the work to decide whether the operator is real before deciding whether the operator is right. Authenticity answers the first question. The frameworks answer the second. The order is not negotiable. A buyer who does not believe the operator is real never reads the frameworks.

Transparency

Transparency is naming what others will not. The industry whispers about pricing, and the operator publishes the range. The peer firm hedges on limitations, and the operator admits them inside the proposal. The category avoids declining engagements out loud, and the operator says no in public when the fit is wrong. The pattern is the same in each case. The thing the buyer wonders about and the operator could hide is the thing the operator names anyway.

That principle has a cost, and I learned the size of the cost early. In 2008, my late wife Sylvie wrote a report called Internet Marketing Sins: A Manifesto. The recession had pushed too many operators in our community toward selling things they should not have been selling, and she had been watching the damage from the customer support seat. She was going through chemotherapy at the time. The verbal fight with bad actors had gotten too costly, so she wrote the fight down and sent it into the same community we both made our living inside.

The bill arrived fast. We got blacklisted from events. Clients dropped us. Some of the pushback came from people we had worked with for years. The currency we earned back was the one that compounds. Respect from operators who had been waiting for someone to say it. New relationships with buyers who had been looking for someone they could trust. Sylvie’s line, which I still carry, was simple. Make money at the service of others, not at the expense of others. The transparency principle that anchors one third of EAT 2.0 was lived before it was named. The manifesto was 2008. The framework arrived later. The principle was already in the room.

In a fractional or expert practice, the same principle compounds through small repeated acts. The case study published with the parts that did not work alongside the parts that did. The result reported with the methodology underneath it, not just the headline number. The credit shared with the team or the predecessor whose work made the result possible. The buyer reading the pattern across a year of those acts is the buyer who decides to call. Each act looks small in isolation. The pattern is what the reader is reading.

True thought leadership

Most operators use the term thought leadership to describe a thinner version of it. How-to content with mild opinion attached. The operator pulls from the same conventional wisdom every peer pulls from, adds a personal anecdote, and publishes the result under the leadership label.

That is leadership of thoughts the field already had. Real thought leadership produces something the field did not have before the operator brought it. Three forms it can take.

Unique research. The operator surveys their own list. Runs an original poll. Publishes the data with their own interpretation rather than citing someone else’s. Google’s Quality Rater Guidelines specifically reward unique research, because the engine is trying to elevate sources that produce the material the field is citing rather than sources that are doing the citing. The operator who runs the research earns the citation tail behind it.

A unique point of view. A perspective that differs from the consensus and is defended on the merits. Sylvie’s manifesto was a unique point of view, defended in plain language, at cost. Cost is what tells the reader the position is real. A free opinion is an opinion no one is paying for. A position the operator can name a price for has weight no free opinion carries.

Named frameworks. Power Positioning. FAME. OATH. QUEST. FORCEPS. The Bullseye Method. Revenue Architecture. EAT 2.0 itself. Each one began as a private way I made sense of work I was doing, and turned into a unit of authority other people quote, teach, and pass on. The framework becomes a carrier of authority once it has a name the field can repeat, and the act of giving it a name is what brandifying produces. The framework gets to do the spreading the operator’s own time cannot.

The three forms compound on each other. Unique research is the kind of thing readers cite. A unique point of view is the kind of thing readers defend. A coined framework is the kind of thing readers teach. Each act of citation, defense, and teaching pushes the operator’s authority into rooms the operator’s calendar never reaches.

The AI irony

The era of AI-generated content is also the era of the highest-value human signal underneath the content. The machine is closing the gap on every part of the work it can imitate. The parts a person has to bring are the parts the market is now paying a premium for.

The reader, the buyer, and the algorithm itself are converging on the same demand. Prove there is a person here. Prove the experience under the page is lived experience. Prove the position is one a real human will defend at cost. Three audiences asking the same question in three different voices, and the operator who answers compounds.

EAT 1.0 measures the surface. EAT 2.0 carries the human layer underneath. The operators who treat the two as a stack rather than a substitution are the operators whose authority compounds across the AI era. The framework I first wrote about as the humanization strategy has a sharper name now, and the name is the move EAT 1.0 was never asked to make.

Empathy lives in the work or it does not. Authenticity is visible before the reader reaches the first framework. Transparency costs what it costs, and the cost is the currency the relationship is built in.

Authority compounds on the layer AI cannot flatten. That layer is EAT 2.0.

---

Frequently Asked Questions