AI content gone wrong

The rise and reality check of Grokipedia

May 05, 2026

👋 Hey, I’m George Chasiotis. Welcome to GrowthWaves, your weekly dose of B2B growth insights—featuring powerful case studies, emerging trends, and unconventional strategies you won’t find anywhere else.

This note is brought to you by Minuttia.

Most people talking about AI search are still guessing.

We’re not.

In May, we’re running a 2-day intensive AEO course for people who want to understand what actually works and how to apply it.

This isn’t theory. It’s the exact playbook we use:

How AI search really works (and how it differs from SEO)
How to build visibility across AI platforms
Where AI helps and where it kills quality
The workflows, templates, and tools behind our processes

The course is live, hands-on, and capped at 30 people.

If you want to move past noise and build real visibility in AI search, this is for you.

Join the cohort

Six months ago, xAI launched Grokipedia.

An AI-generated encyclopedia built to rival Wikipedia.

Powered by Grok, seeded from Wikipedia’s top one million articles, and scaled to over six million entries in half a year.

For a while, it looked like the experiment was working. Traffic surged. Google indexed hundreds of thousands of pages. Keywords piled up.

By January 2026, Grokipedia was pulling in over 1.6 million organic visits per month.

Then it all reversed.

I have been tracking this story since early in its run. I wrote about it twice on LinkedIn (here and here) when the growth numbers were still staggering.

Now the full picture is coming into focus, and I think there are lessons here that every marketing team publishing AI content at scale needs to hear.

The Numbers

Grokipedia launched on October 27, 2025, with 885,279 articles. Within six months, that number crossed six million.

To put that in context, English Wikipedia has roughly 7.2 million articles built over 25 years. Grokipedia got to ~87% of that corpus in 180 days.

The traffic story was just as dramatic.

Ahrefs data showed the site ranking for close to 1.9 million organic keywords at its peak, and was bringing in 1.6 to 1.7 million visits per month.

The growth was real.

But growth without quality has a shelf life.

How the Content Gets Made

Grokipedia is not a crowdsourced encyclopedia. It is a machine-generated one.

The process starts with Wikipedia.

xAI’s team instructed Grok to compile the top one million Wikipedia articles and then “add, modify, and delete.”

That means research the rest of the internet, correct what Wikipedia got wrong, fix mistakes, and add context.

Some articles were republished nearly verbatim from Wikipedia, accompanied by a disclaimer noting that the content was adapted under a Creative Commons license.

Others were completely rewritten by Grok from scratch.

Here’s what the process looks like:

The editorial model is centralized.

There are no volunteer editors in the Wikipedia sense.

Users can suggest edits and report errors, but all changes go through Grok’s AI review system. The AI decides what to accept, reject or modify.

This is a completely different approach to building an encyclopedia.

Wikipedia’s model relies on distributed human judgment.

Author’s Note: I’m not saying that’s right, but it is what makes Wikipedia what it is today.

Grokipedia’s model relies on a single AI system making editorial decisions at scale.

The result is an encyclopedia where every article reads like it was written by the same author, which in this case is AI without a human-in-the-loop.

The Indexation Gap

Here is where the data gets interesting.

Grokipedia has over six million articles. But when you run a site:grokipedia.com search on Google, the result shows approximately 1.5 million indexed pages.

That is a 75% indexation gap. Three out of every four articles on Grokipedia are not in Google’s index.

When I wrote about this on LinkedIn back in February, the gap was 65.11%.

It has gotten worse, not better.

Google is indexing a smaller and smaller share of the total article count.

Ahrefs crawl data tells a similar story.

The tool shows 1.57 million crawled pages for grokipedia.com.

Of those, ~282K return a 404 status code.

That is 17.9% of all crawled URLs leading nowhere.

I’d guess that Google is either a) refusing to crawl the site, or b) it is crawling the pages and choosing not to index them, or a combination of both.

Which, in either case, means Google has seen the content and decided it does not meet the bar.

The Decline

The peak came in January 2026. After that, the numbers collapsed.

By April 2026, Ahrefs shows Grokipedia ranking for 326,000 organic keywords. Down from 1.9 million at the peak. That is an 83% decline in three months.

Organic traffic dropped to 613,000 visits. Still a great number, but less than 40% of the January peak.

The top-3 keyword positions tell the sharpest story. Grokipedia held 24,074 keywords in positions one through three in February.

By April, that number was 7,081. A 71% drop in the keywords that actually drive clicks.

On February 6, 2026, SEO consultant Glenn Gabe flagged that Grokipedia’s Google visibility had fallen off a cliff.

The timing lined up with what the data was already showing.

The Search Experience Problem

I want to show you something that captures the deeper issue.

Go to Grokipedia and search for “what is a royal advisor.” The site returns over 10,000 results.

None of them on the first page are about royal advisors.

While there’s a page that’s dedicated to that topic:

This is not a minor glitch. It reveals a structural problem.

When you generate six million articles with AI, you can cover a lot of ground. But covering ground is not the same as covering it well.

The search experience on Grokipedia surfaces articles that contain matching words but miss the intent entirely.

This is what happens when volume replaces editorial judgment.

You get breadth without depth, coverage without comprehension.

What This Tells Us About AI Content at Scale

I wrote about AI content strategy in my note on Content Engineering.

In that piece, I outlined four principles for content engineering and for safely deploying AI content strategies.

Every one of those principles applies here. Grokipedia violated most of them.

The content was generated without meaningful quality gates
The editorial review was automated rather than human-supervised
Scale was prioritized over substance
And the feedback loop between search performance and content quality was either too slow or nonexistent

The 99.8% informational keyword split is telling. Out of 326,000 keywords Grokipedia ranks for today, 325,300 are informational queries.

Almost zero commercial or transactional intent.

This is a site built for volume, but not necessarily for value.

My Take

I give xAI credit for the ambition. Building an encyclopedia from scratch and reaching six million articles in six months is a technical achievement. The speed alone is worth studying.

But technical achievement is not the same as content quality. And content quality is what search engines reward over time.

Grokipedia is the clearest case study we have of what happens when AI content meets reality at scale. The initial spike was real.

Google indexed the pages. Traffic surged. Keywords multiplied.

Then the correction came.

Google saw millions of pages generated by a single AI system. It indexed a fraction. It ranked even fewer.

Over three months, it pulled back on the rankings it had initially given.

This is the pattern I keep coming back to. AI can generate content at any scale you want.

The constraint is not production. The constraint is quality.

And the market for quality is still governed by human judgment, both from search engines and from the people reading the content.

Here is what I tell every team I work with:

Even if you are a tech company with unlimited resources, these dynamics are going to catch up with you. At some point, they will redefine what your content strategy can achieve. You have to be responsible about how you deploy AI content at scale.

Grokipedia had one of the most powerful AI systems in the world behind it. It had the infrastructure, the data, the compute. It still hit the wall.

The wall is not technical. The wall is human.

Final Thoughts

The Grokipedia story is not over. The site still pulls 613,000 organic visits per month. It still has massive compute and resources behind it.

And, for what it’s worth, it could recover at some point in the future.

But right now, Grokipedia is a six-million-article case study in what happens when you scale production faster than you scale quality.

The four principles from my Content Engineering note were not theoretical. They were written because I keep seeing this pattern play out. Grokipedia is the largest example yet.

Build with AI. Scale with AI. But do not hand AI the editorial keys and walk away.

That is the lesson. And it applies to every company running AI content right now, not just the ones with billion-dollar budgets.

Thank you for reading today’s note, and see you again next week.

Research Disclaimers and Limitations

GrowthWaves and its author are not sponsored by or compensated by any company mentioned in this note. This is independent editorial analysis and does not constitute investment, financial, or legal advice. The author may have relationships with, work with, or hold equity in companies referenced; however, no content in this piece was influenced, commissioned, or incentivized by any such relationship. AI tools were used as a research assistant in the preparation of this piece. All claims are sourced and linked throughout.

GrowthWaves by George Chasiotis

Discussion about this post

Ready for more?