In Saigon, at a hotel ballroom filled with sleep-deprived SEO consultants, Pavl Kilico walked on stage and delivered a two-hour lecture on originality. Not the “be yourself” kind of originality, but the kind Google measures when it decides whether your website deserves to exist. The crowd laughed, nodded, and took furious notes, because nothing is funnier than realizing your career depends on tricking a math problem into believing you’re a real person.
Pavl is from Estonia, which has fewer people than some American suburbs. Yet he has become something of a folk hero in the holistic SEO community — a digital guild of 40,000 members swapping notes on how to survive the next algorithm update. He spends winters in Thailand and Vietnam, summers in Europe, and all year explaining to SaaS companies why their shiny dashboards don’t matter if Google thinks a robot wrote their content.
This is the paradox of modern search: we live in an age where every piece of text could be AI-generated, yet the most valuable thing online is proof that a human touched it.
When originality became a ranking factor
For years, search engine optimization was about speed. Around 2019, Google rewarded websites that were constantly updated. If you flood the web with 3,000 articles a day, you could win. It was an arms race of quantity. Then large language models arrived, and suddenly, anyone could do that. Which meant Google had to change the rules. Momentum no longer mattered; originality became the new currency.
This is why forums are back. Reddit threads, Stack Overflow questions, even Quora rants — they rank because they look alive. They contain disagreement, bad grammar, jokes, and digressions. They smell like humans. In contrast, content farms — those infinite walls of blog posts with stock photos and keywords — look as dead as they feel. Kilico put it bluntly: “A content site without a function is just a wall with text. And a wall doesn’t rank.”
The originality arms race
The search industry now operates on the assumption that every two years, Google engineers pick a new way to tell quality from spam. In 2019, the focus was on speed, in 2021, on freshness, and by 2025, on originality. Originality is not plagiarism detection. That’s the kindergarten level. It’s about signatures — tiny statistical tells in the way sentences are built, the HTML is structured, even the placement of interactive elements like sliders and checkboxes.
There’s an entire field called stylometry that grew out of a dispute over who wrote 13 mysterious pages of the U.S. Constitution. Historians used word distance and frequency analysis to decide it was Alexander Hamilton, not Thomas Jefferson. Google uses similar techniques to determine whether you wrote your product descriptions or if you asked ChatGPT to do it during coffee breaks. And this matters because the entire business of SaaS tools is built on misunderstanding this.
The SaaS founder’s curse
Here’s how it usually goes:
- A developer builds a tool for themselves. It works.
- They realize they can sell it.
- They add a dashboard, copy their competitors’ features, and suddenly they’re just another average SaaS with a different logo.
The market is flooded with these tools, many of them promising to “beat Google” with AI content generation. Most people are unaware that Google holds patents specifically for detecting machine-generated text at the word level.
Kilico has consulted with SaaS companies that wanted to measure “originality” but had no idea what it meant. They thought it was about plagiarism. They didn’t realize Google might be analyzing the way users click a slider, or how often synonyms co-occur in sentences, or whether a page layout resembles known spam templates. As he put it: “In 99 percent of cases, they don’t know what they’re dealing with.”
Forums don’t cannibalize
One of Kilico’s favorite examples is the difference between websites and forums. If you publish two similar pages on a blog, they cannibalize each other — Google can’t decide which one should rank, so they both sink. However, if two similar threads exist in a forum, they are consolidated. They actually strengthen each other.
Why? Because forums are living systems. Multiple voices reinforce authority. A forum with ten threads about the same allergy medicine looks like a conversation. A website with ten articles about it seems like spam. This is why Kilico has been creating “exact match subreddits” — little communities named after dictionary words like “allergy”. It’s the new version of buying exact match domains in 2005. He’s betting Google will prefer messy human chatter over pristine AI prose.
The Microsoft study: Five humans who can smell robots
In one Microsoft study, researchers hired 100 freelancers to guess whether AI wrote a text. Most failed miserably. But five of them scored over 98 percent accuracy. What was their secret? They used AI daily. They recognized its tics the way a poker player recognizes a nervous twitch. They knew the default phrases AI leans on, the way it over-explains, the way it piles multiple ideas into a single sentence.
Microsoft then codified those patterns into a detector. Imagine: five bored freelancers in sweatpants quietly shaped the future of copyright law and education policy. And yes, Microsoft spent millions on this research — not for fun, but because detecting AI content saves them money. Crawling and indexing AI sludge is expensive. If they can filter it out early, they protect their search engine, their training datasets, and maybe the entire business model of Bing.
Watermarks and the politics of detection
Of course, there’s a more straightforward way to detect AI: watermark it at the source.
Google has a system called SynthID. It tweaks the probabilities of word choice so the output carries a hidden statistical signature. To the reader, it looks normal, but to the algorithm, it glows neon green.
In theory, this solves the problem. In practice, it introduces new ones:
- What if people copy-paste AI text and edit it? The watermark gets blurred.
- What if students unknowingly echo watermarked text? They get accused of cheating.
- What if companies don’t want to share their watermark keys with rivals? The detection ecosystem breaks down.
And then there’s the hilarious side effect: once people start reading watermarked AI text, they unconsciously adopt its phrasing. Which means humans began writing in a style characterized by watermarks. The system designed to separate humans from machines ultimately turns humans into machines.
Why originality is about more than words
Let’s return to that credit card application form. Version A has a slider where users can select their budget. Version B replaces it with checkboxes. Exact text, same images, different interactivity. Which is more original?
Kilico argues that originality isn’t just about vocabulary. It’s about responsiveness. The slider creates a unique experience — the user is literally generating their own sentence: “I want a credit card with a $5,000 limit.” Google can parse that. It verbalizes the slider range into text tokens, creating relevance for every number between 5,000 and 100,000. Suddenly, your page covers hundreds of possible queries without adding a single paragraph.
Originality can be coded into HTML. A dropdown menu, a tab structure, or even a well-labeled button can be a signal of human effort.
Clusters, representatives, and the curse of sameness
Here’s the nightmare scenario for small publishers:
- AI models generate millions of near-identical articles.
- Google clusters them by vocabulary similarity.
- Only one “representative” survives. The rest are buried.
Who gets chosen as representative? Typically, it’s a big brand, e-commerce site, or SaaS platform. The long tail dies in the cluster. This is why originality matters. It’s the only way to escape being absorbed into the cluster. If your article is just another “what are the symptoms of X,” you disappear. By incorporating real experiences, alternative perspectives, or unique interactivity, you become stand out enough to be chosen.
Safe answers and the bulletproof document
One of Kilico’s frameworks is the idea of “safe answers.”
Instead of committing to one version of the truth, cover all the perspectives:
- Facts: “According to clinical trials, CBD oil shows limited evidence for anxiety relief.”
- Experience: “Users on forums report improved sleep after using CBD oil.”
- Perspective: “Some regulators argue CBD is overhyped and under-tested.”
By splitting content into these layers, you bulletproof it against consensus shifts. If public opinion shifts, part of your article will remain accurate. This also happens to mirror how Google processes queries: factual, opinion, or experiential. If your content provides all three, you’re always relevant to at least one interpretation.
The SEO writer’s dilemma
Picture an SEO writer in 2025. Their manager asks them to “make it original.” They open a content brief that reads like a PhD thesis, covering a range of technical terms, including synonyms, co-occurrence distances, pixel spacing, trigram coverage, HTML annotations, and human signatures.
They stare at the screen, wondering if maybe plumbing school would have been a better career move. Meanwhile, Google’s AI Overview system is spitting out different answers to nearly identical questions: “How does hosting work?” “How does hosting function?” “What does hosting do?” Each phrasing triggers a different summary.
The writer realizes their job is to anticipate not just keywords, but every possible way a human might phrase a question, then bake all of those variations into a single article without making it unreadable. This is what originality looks like now: a war of synonyms, annotations, and perspective markers.
Cultural consequences: The trust paradox
Here’s the irony: AI makes writing easier than ever, but because everyone uses it, trust is more complex than ever. If you land on a medical clinic’s website and the content feels AI-generated, you won’t let them cut into your body. If your dentist presents a treatment plan and says: “ChatGPT said this should work,” you’ll likely run.
That’s why Google now rewards human signals even when they’re wrong. A messy forum thread with incorrect answers outranks a perfectly accurate AI article. Because at least the humans believed what they were saying. It’s a strange new hierarchy: truth matters less than proof of humanity.
Originality as a moving target
The lesson from Pavl Kilico’s talk is not that originality is solved, but that it is perpetually redefined.
- In 2019, originality meant “publish faster.”
- In 2021, it meant “update more.”
- In 2025, it means “show human effort.”
- In 2027, it might mean “embed biometric proof you sweated while writing this.”
Every time AI closes the gap, Google invents a new filter. Watermarks, stylometry, HTML annotations, cluster representatives — each is a new hoop for publishers to jump through. And each time, an industry of tools springs up promising to automate the hoop-jumping. Most fail. The real winners are the researchers who quietly sell their detection methods to Microsoft or Google.
Why this matters beyond SEO
If this all sounds like inside baseball for search nerds, consider the broader stakes:
- Copyright law: Who owns a book if AI can mimic an author’s style perfectly?
- Education: How do professors grade essays if detectors falsely flag non-native speakers?
- Politics: What happens when AI floods the web with plausible but fake policy papers?
The systems that decide originality online will shape not just who gets web traffic, but who gets to speak, publish, and be trusted.
Conclusion: the invisible teacher grading our homework
Search engines are no longer just matching words. They’re grading us on effort. Did we add perspective? Did we structure our HTML with intention? Did we build an interface that anticipates user needs? It’s like having a strict English teacher who also happens to run the internet. And every time we find a way to cheat, she invents a new rubric.
Pavl Kilico ended his talk with a simple suggestion: join a community. Share notes. Because no one can keep up alone, that’s the fundamental cultural shift: SEO is no longer about tricks. It’s about collective survival in an ecosystem where the most important thing you can prove is that you are, in fact, human.
Follow us on X, Facebook, or Pinterest
