Since the beginning of SEO, or really from the beginning of search itself, we have always had one major issue: the gaming of the system.

The Era of Keyword Stuffing

In the very earliest days, before links even mattered, the game was keyword stuffing. Search engines ranked pages largely on how many times a word appeared on them. SEOs crammed the same keyword into titles, meta descriptions, alt text, headings, and body copy, sometimes dozens of times per page. Some even hid extra keywords in white text on white backgrounds or tiny font sizes.

It worked brilliantly for a time. The top 10 results were often the pages that simply repeated the target phrase the most, not the ones that actually answered the user's question best. Google eventually fought back with the Florida Update in November 2003, which specifically targeted keyword stuffing and over-optimised affiliate sites.

Link weight changed dramatically after that. At first, links were king, more or less regardless of where they came from. If you got a .gov or .edu link, those were holy grails. The higher the PageRank of the linking page, the better the outcome.

Links were (and still are) highly gameable, so Google started categorising them. A link from the same sector with keyword-rich anchor text became gold, and of course, that got gamed too.

Explore the SEO Founders’ User Guide — a distilled resource built from real-world experience across the SEO industry. Cut through the noise and focus on what actually works.

Then they moved to the nofollow attribute in January 2005, hoping to neutralise paid links and comment spam (as announced in Google's official blog post "Preventing comment spam"). SEOs simply built nofollow networks and reciprocal schemes instead.

So Google rolled out Penguin in April 2012, the first major algorithmic hammer aimed squarely at unnatural link patterns, over-optimised anchor text, and private blog networks (PBNs). It impacted around 3.1% of English search queries.

But the game didn't stop. It just evolved. SEOs moved to "guest posting," "niche edits," "expired domain PBNs," and later "parasite SEO" on high-authority platforms. Google responded with more link spam updates (including July 2021 and December 2022) and brought in SpamBrain, their AI-based spam-prevention system designed to automatically detect and nullify unnatural links at scale.

Content Farms and the Panda Update

Whilst all that was happening with links, keyword page stuffing quietly evolved into something more sophisticated. After the early crackdowns, SEOs stopped the obvious repetition and moved to "content optimisation" at scale. They created thin, formulaic pages stuffed with just enough keywords to rank, then spun hundreds of near-duplicate variations.

This exploded into the content farm era, most famously Demand Media and eHow, which were pumping out thousands of low-effort articles a day purely to rank. Google hit back hard with the Panda update in February 2011, targeting low-quality, keyword-heavy, low-value content.

But again, the game simply shifted: SEOs learned to write longer pages, add LSI keywords, and later use AI tools to generate "helpful-looking" filler at a massive scale. Google responded with the Helpful Content Updates starting in August 2022 (with further iterations in December 2022, September 2023, and into 2025), which were designed to reward content written for people rather than search engines.

The Uncomfortable Truth

Here's the uncomfortable truth about every one of these tactics: keyword stuffing, link gaming, page stuffing, content farms, parasite SEO, whatever the current flavour is. They all worked. For a while, at least.

The top 10 results were consistently the sites that mastered the current game, not necessarily the ones with the best content, the best product, or the most genuine expertise. Google would announce a new signal to fight manipulation, the SEO community would adapt within months, and the cycle would repeat. Every single time.

And Now We Have EEAT

Google tells us the answer is no longer just keywords or links. It's Experience, Expertise, Authoritativeness, and Trustworthiness. They say this is the new north star, especially for YMYL topics like medical and aesthetics. They even have human quality raters evaluating pages against it.

John Mueller has repeatedly emphasised that "EEAT is not something you can add to a website. That's not how it works."

The Reputational Problem

The derogatory attitude from a plethora of SEOs about EEAT is truly stunning. Respected SEO professional David Quaid recently captured the frustration perfectly whilst discussing the New York Times investigation into AI Overviews:

"Reading an article about the NYT and AIO being wrong and E-E-A-T and its like.... its like Google can't detect E-E-A-T. Because it's a nice idea that people 'think about EEAT' and as Google says: you can't 'add it to your site' because making claims does not equal EEAT."

He is not wrong. And this is the crux of the problem, and why EEAT already carries a reputational dent in the SEO community.

The Gap Lily Ray Identified

Lily Ray's recent piece (April 10, 2026) is right to call out the gap. Her examples of high-intent commercial queries, "best AI chatbot", "best SEO tools", "best project management software" and many others, show AI Overviews frequently surfacing low-effort, self-promotional listicles that any SEO can produce quickly.

These are exactly the kind of thin, manufactured "Top 10" pages that should fail a proper EEAT evaluation, yet they get cited and summarised as authoritative answers.

The Reality: AI Is Still in Its Infancy

If we take a pragmatic approach, we have to acknowledge that AI, at least in its current form as used in search Overviews and generative answers, is still an infant. We are essentially back in the keyword-stuffing days of 2003.

The algorithm is underdeveloped, the "AI brain" is heavily randomised by token usage and training data patterns, and it cannot truly reason. Instead, it skims surface signals, favours frequency and recency over depth, and often elevates content that looks plausible rather than content that is genuinely authoritative.

AI might stand for Artificial Intelligence, but right now it is anything but intelligent in the way Google claims when it talks about EEAT.

Is EEAT Itself Biased?

This brings us to a deeper and more uncomfortable question: is EEAT itself biased, and is that part of the problem?

Yes, at least in practice. If you are a doctor, dentist, scientist, or any key academic with deep, verifiable tentacles into natural EEAT (peer-reviewed papers, medical council registration, hospital privileges, conference speaking, long-term clinical outcomes), it is reasonable to hypothesise that AI might one day learn to recognise and properly weigh those signals. The raw material is rich and externally verifiable.

But this represents only a tiny sliver of society.

What About Everyone Else?

What about the thousands of other professions that do not have the same natural, high-signal EEAT footprint? Small business owners, consultants, tradespeople, creative professionals, local service providers, the vast majority of the economy.

For them, there is no equivalent "external validation machine." They are left trying to simulate EEAT through on-page claims, testimonials, or clever content, which brings us straight back to the original issues of gaming and manipulation.

The Cycle Never Ends

We end up with the same problem we have always had: results that are gamed. Even today, with all its sophistication, Google still serves up ten results that are often more about SEO mastery than genuine quality. Its understanding of context remains loose enough that manipulation persists.

If traditional search, with decades of refinement and human raters, still struggles with this, how exactly is an immature AI system supposed to get past it?

That is the real question Lily Ray's piece forces us to confront. The history of search shows a consistent pattern: new quality signals are introduced with the best intentions, yet they quickly become just another lever in the ongoing game of optimisation versus genuine value.

The Pragmatic Path Forward

The uncomfortable reality is that we may need to accept that no single framework, whether links, content signals, or EEAT, can fully solve the problem whilst the underlying systems (both traditional search and AI) remain imperfect at distinguishing signal from noise at scale.

For now, the pragmatic path for most sites, especially in competitive or YMYL spaces, is to focus on building real substance that serves users first, whilst recognising that AI's ability to reliably reward it is still very much a work in progress.

What do you think? Is EEAT ultimately a helpful guiding principle, or just the latest chapter in an endless cycle? The answer probably lies somewhere in between, but the conversation Lily Ray started is one worth continuing.