The short version: I ran my own website as an experiment in publishing AI-assisted content at volume. In 45 days I published 569 posts. Google indexed about 5 percent of them. Not because the posts were thin, they averaged over 2,000 words, but because a website earns the right to have its pages shown, and that right is rationed by how trusted your site is. Publishing faster does not buy you more of it. This is what actually happened, with the real numbers, and what I am doing next.
I need to tell you where I was first
In 2019 I was earning well into six figures from this business. By 2021 I was rebuilding it from nothing, in public, after five years I would not wish on anyone. So when AI writing tools got good enough to publish real articles at scale, I did not approach them as a gimmick. I approached them as a lifeline. If I could publish a lot of genuinely useful content, quickly, I could rebuild my traffic and my business at the same time. So I built a machine to do exactly that.
What the machine did
It wrote and published blog posts on autopilot, around a dozen a day, every day. Real posts. Each one over 1,700 words, structured properly, on topics my audience actually searches for. Work from home, freelancing, AI for small business, making money online. Over 45 days it produced 569 of them.
I assumed the hard part was writing the content. I was wrong. The hard part is something nobody selling you an AI content tool will mention.
Then I looked at the data, and it stopped me cold
I went into Google Search Console and checked a simple thing: of those 569 posts, how many is Google actually showing to anyone?
The answer was 5 percent.
Twenty-eight of them. The other 541 were invisible. Not penalised, not banned. Just sitting there, never shown to a single searcher. Months of an automated machine running flat out, and 95 percent of its output might as well not exist.
My first instinct was the obvious one: the content must not be good enough. So I checked the length. The median post was 2,028 words. I checked whether the longer ones did better. They did not. My 1,500-word posts and my 2,500-word posts got shown at exactly the same rate. Length was not the problem.
Here is the thing that actually explains it
Think of Google as an enormous library, and every post you publish as a book. For anyone to ever find your book, a librarian has to put it on a shelf first. Until then it sits in a back room nobody can enter.
The catch is this: the librarian only shelves so many books from each author, and how many depends on how well known and trusted that author is. A famous author gets shelf space for everything they write. An unknown one gets a few books on the shelf and the rest left in the back.
My site is trusted at a middling level. So Google shelves maybe one of the dozen books I hand it each day, and quietly leaves the rest in the back room. Writing more books does not change how many the librarian will shelve. It just fills the back room faster.
I proved this to myself with one more number. When I was publishing slowly, earlier in the year, around 75 percent of my posts got shelved. The moment I ramped up to a dozen a day, that dropped to 7 percent. Same machine. Same writing. The only thing that changed was the speed. The faster I published, the smaller the share that got seen.
The honest lesson, the one I wish someone had told me
You cannot publish your way to traffic on volume alone. The number of pages a site can get shown is capped by how trusted that site is, and you raise trust by getting other respected websites to mention and link to you, not by writing more. Volume and trust have to grow together. Pour volume into a site with low trust and you do not build an asset, you build a back room full of invisible pages, and you can actually make the whole site look worse in Google's eyes.
This is not an argument against AI content. I am still publishing daily and I am not stopping. It is an argument against the lie that volume is a strategy by itself. It is half a strategy. The other half is trust, and the other half is the hard part, because it cannot be automated as easily.
What I am doing about it
Three things. I made each post genuinely more original, the kind of first-hand, specific, opinionated writing that earns its place instead of repeating what is already online. I am building my content into tight topic clusters so the pages support each other instead of competing. And I am going after trust directly, the slow, human work of getting respected sites to point at mine. The first two improve the odds for each page. The third is the only thing that raises the ceiling for all of them.
I will report back with the numbers, because that is the whole point of doing this in public.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
The 11-Post Cluster That Ranked: What Made It Different From the Other 558
Out of those 569 posts, I found one tight cluster of 11 articles that cracked page one within 60 days. They were all targeting long-tail queries around a single B2B SaaS category, and every single one of them had something the rest of the batch did not: I had spent 20 minutes personally editing each one to add a specific data point pulled from an industry report that was less than 90 days old. Not rewritten, not restructured. Just one injected paragraph with a real number and a citation the AI had no access to. That is it. That was the variable.
The other 558 posts were cleaner in some ways. Better formatted, more consistent tone, faster to produce. But they were built entirely from what the AI already knew, which means they were built from the same inputs as every competing article on the same topics. Google does not need a 559th version of information it already has indexed from 2022. The 11 that ranked gave it something it could not already retrieve. A fresh stat, a current source, a timestamp it could trust. I now call this the "one true thing" rule before any post goes live: what is in here that did not exist on the internet six months ago?
If you are running AI content at scale and wondering why Google is suppressing most of it, run this audit on your own batch:
- Pull your top 10 indexing failures and paste the intro into a search query
- Count how many near-identical paragraphs appear in existing ranked results
- Check whether your post contains any data, quote, or observation published after the AI model's training cutoff
- If the answer to that last point is no, the post is not a content asset yet, it is a draft
The honest number is this: adding that one paragraph of fresh, sourced specificity took me an average of 18 minutes per post. On 569 posts that would have been 170 hours of extra work, which obviously kills the economics of the whole experiment. But on a targeted cluster of 50 posts around your highest-value topics, that is 15 hours of editing that could realistically move the needle. The mistake I made was treating all 569 posts as equal when volume itself was never the point. Precision on the posts that matter is the point.
Frequently asked questions
Why are most of my blog posts not showing up on Google?
Most likely they are not indexed, which means Google has not added them to the pages it is willing to show. Google rations how many pages it indexes from each site based on how trusted the site is. If you publish faster than your trust level supports, the surplus pages stay invisible no matter how good they are.
Does writing longer blog posts help them get indexed?
Not on its own. In my own data, 1,500-word and 2,500-word posts were indexed at the same low rate. Length only helps when it adds genuinely new information. A long post that repeats what is already online does no better than a short one.
How do I get more of my content indexed by Google?
Raise your site's authority by earning links and mentions from other respected websites, link your own pages together into clear topic clusters, and make each page genuinely original. Authority is what lifts the ceiling on how many pages Google will show.
Is publishing AI content at volume a bad idea?
No, but volume alone is not a strategy. AI lets you publish a lot, fast. Whether those pages get seen depends on your site's trust and the originality of the content. Pair volume with authority-building and original writing, or most of it stays invisible.
Want this done for you? See running your marketing operations with AI.
For the bigger picture, see my full guide to AI marketing.
Related reading: How to Find Your First Freelance Writing Client (Without Sending 200 Cold Emails Into the Void) and How to Keep Your Brand Voice When You Scale With AI.
