I handed my content marketing to agents. 6 months later organic traffic is 34x.

That's Search Console for the last 12 months on a site I've run for years. Everything flat on the left is before February.

I'm solo on this. I haven't written, researched or edited a single one of the articles that did it.

Here's what the six months actually looked like, including the two months I wasted.

What the setup is

The site publishes 200 articles a day now. It put out 200 today, and there are just over 23,000 live.

I've been doing SEO for 22 years, so the pipeline isn't anything clever. It's what I would do by hand for one article: pull market research from Google Ads data, settle on a keyword, fetch the top 10 results and read what they cover, then write something that covers all of it and adds what none of them bothered with.

The only change is who does it. That work now runs as agents, end to end.

The first two months were a write-off

Of everything published in April, 53% never got a single impression, even after 60 days. Researched, covered, shipped, and more than half of it might as well not exist.

My first guess was that Google could tell a machine wrote it. Wrong. I pulled index status on the dead pages and 66% were indexed normally. Not rejected. Indexed, and nobody was searching for them.

So I checked all 1,412 dead pages for any adjacent keyword doing 100 to 2,000 searches a month. Found one for 158 of them. Eleven percent. The rest were sitting in a vacuum.

The pipeline was fine. I'd pointed it at keywords that were too big. Cover the top 10 for a competitive term and you land at position 11, and position 11 gets you nothing. Doing that a thousand times doesn't change the arithmetic.

At 1,000 to 5,000 searches a month, my median position is 14.4 and only 34% of those queries reach the top 10. At 50 to 100 searches a month, median is 8.6, 68% are in the top 10 and CTR is 2.90%, the best of any band.

From June I kept every step and only moved the target down. Zero-impression rate went 53% in April, 35% in May, 25% in June. Same agents, same pipeline.

What that does to the business

With February as 1:

Clicks from search, 33.9x. Ad revenue, 33.6x. People on the site, 20.0x. Pageviews, only 3.0x.

Revenue tracked clicks almost exactly, so the value of a click held up while the volume went up 34 times. Nothing got diluted. September had passed the whole of August by the 16th.

The pageview line is the one founders ask about. People grew 20x and pages grew 3x, because someone arriving on a long tail query reads the one article they came for and leaves. The site went from 12.4 pages per person in February to 1.9 in August. This audience does not browse. That was fine, because each article answers exactly one question and the monetisation on it is relevant to that question.

In February, 209 pages on the whole site got a click, the top 10 took 70.9%, and one page by itself, the one people land on by typing the site's name, was 54% of everything. All head, no body. By August, 6,839 pages got a click, the top 10 took 10.1%, and the biggest single page was 2%. If any one page vanished tomorrow I could not tell from the numbers.

The traffic number is not really the interesting bit to me. Nothing in it is load-bearing any more, which is a very different position to be in than having one page carry the site.

The thing I wasn't expecting

Up to March this site had never once had a link exchange request. Not one, in years.

Once the blog section existed and the articles piled up, they started coming in almost daily. I've never done outreach and I've never run a link building campaign. Adding a blog was the whole intervention.

And it's not who I'd have guessed. Companies much bigger than mine, listed companies, law and accounting firms, the kind of organisation that would not have taken my call a year ago. Almost nothing from the link farm end. Whatever a stocked blog signals, the people whose job is to vet a site before linking to it are reading it as legitimacy.

Half of this is deletion

Nobody talks about this part and I think it matters more than the publishing.

28 days with zero impressions and a script redirects the topic. 56 days and it sets the page noindex and drops it from the sitemap. That runs every morning without me. On August 19 it removed 1,601 articles in one pass and I found out from the log.

If you only ever add, the share of your site that is dead climbs forever, and that ratio is the thing a spam system can actually see. Raw volume tells it very little.

What my job is now

Choosing which tier to aim at, and signing off on what gets cut, is not automated and it isn't going to be. That part has to be argued out against the numbers, and an agent on its own has no stake in the answer.

In practice: every morning the agents report in. What got impressions. What has gone 28 days with nothing. What the retarget queue is holding, which is 2,418 articles as I write this. I read that against the KPIs, we talk through what it means, and I decide what the next batch is about.

So the job is one decision a day, off a report I did not compile, about work I am not going to do. It took a while to stop feeling like I was skiving.

Next thing I want to fix is that retarget queue. 2,418 is more than I am getting through and it has been growing for weeks, which means there are a few hundred articles sitting at zero that could probably be pointed somewhere useful. I have not worked out how to make that decision faster without handing it over, and I do not want to hand it over.