I take no pleasure in saying this – niche content websites are toast.
I’ve spent well over 1,000 hours of my life creating & growing sites like these as businesses, and I’m frustrated by the conclusions I draw here.
There remain many people making money in the niche content site business, but I’m convinced things will broadly go downhill in the future.
This article explores one of three reasons that I think the content site business model is circling the drain in risk-adjusted return on effort terms. I’ll look into the other two in my next two posts.
The niche SEO content website model goes something like this:
- Start a blog about a topic – typically a narrow one – that people search.
- Write blog posts / articles about said topic.
- SEO to get those articles ranking high on Google’s results page.
- Monetize by selling ad space, affiliate products, or both.
This simply doesn’t work nearly as well as it did 5-10 years ago. I attribute that to three problems that are either new or have gotten worse. I’m covering the first here today.
Few things as costly as cheap content
Millions of people in India & the Philippines getting connected to the internet seems to me like a massive net good for those countries & for humanity writ large, but the consequences of cost-of-living-arbitraged content creation – and now, AI generated content, have been a disaster for both the usability of Google Search & the potential for small publishers to carve out a niche.
I’ll take a second to acknowledge that I’m personally responsible for some portion of this. In the past, I’ve hired a (hardworking & diligent) guy from Lahore to write articles I knew were garbage – but garbage that I knew would rank. I’ve also used GPT-3/3.5/4 to write articles with varying levels of success & quality.
For a time, both tactics worked. I’m not trying to pass judgement on others here.
“The Old Internet” is mostly dead
The approaches above ensure that the internet is drowning in spam, much of it quite well-disguised.
Some have speculated that it’ll become essentially impossible to index the entire web – or any other platform without excellent spam / “proof of humanity” filters – as a result of AI making content creation almost free.
This is (very similar to) what is referred to as “Dead Internet Theory” – a thesis I find more compelling every passing day.
As software engineers know… given a long enough list of things to find, If you can’t index something, you typically just can’t find it at all.
This tsunami of what many would call spam left Google with few options.
Roughly approximated, their approach has been to focus on established & credible content that doesn’t answer specific queries quite as well on average.
This seems to hold true even on more specific long-tail keywords that would historically have been less competitive & easier for small sites to rank for.
Google vs. Publishers
When a measure becomes a target, it ceases to be a useful measure.“Goodhart’s Law” – Economist John Goodhart (Paraphrased)
It’s extremely difficult to accurately rate content in terms of its actual “quality” at the scale of the internet, and Google’s proxies for quality are both weak & manipulatable by its more sophisticated adversaries (i.e., publishers).
This is obviously a problem when your goal is to build a recommendation system for the web.
Google has tried using language models like BERT to get a deeper understanding of the actual content, but that approaches hasn’t been enough.
To better address the low cost of content creation, Google has gradually but continually moved towards backlink authority and “EEAT“, & away from matching the term specifically when it comes to ranking, but those authoritativeness systems have their own problems.
Every credibility system they use can be – and is constantly being – gamed. Buying backlinks is the most obvious example here, but most of SEO falls under this category.
While we’re on the topic, SEO’s might have killed the golden goose (not that there was much of a choice). SEO’s gaining the upper hand over PageRank has lowered trust in Google’s search & forced the big G to look at AI alternatives to the status quo (more on that in a second).
“SEO” as a discipline is primarily an effort in A) writing good content & B) tricking Google’s ranking algorithm by inflating authoritativeness.
Lots of people have gotten quite good at faking the signs of authority, and Google, smart as they are, has run out of levers to pull when it comes to ranking webpages.
Frankly, this makes some amount of sense. There is a finite amount of information available about any given webpage from which to rank it, and it’s simply not enough information to get things “right” every time.
As natural language processing & LLM’s improve, this might get easier, but I’m not holding my breath. Put it all together, & you realize building a search algorithm is genuinely a really hard task. What if Google did something else entirely, though?
How Google is responding with generative AI
They haven’t thrown in the towel. Their upcoming search experience powered by the PALM-2 generative language model promises to answer many questions directly in the SERP – prominently above the list of sites returned from the search.
This will likely solve their customer’s problem of not getting accurate & topical answers, but digital publishers – especially the less-authoritative & smaller sites – will get crushed as generative search handles many queries without the user visiting any other sites besides google.
Niche content sites are not a business I want to be in anymore.
The effort & associated return don’t work in my favor like I once thought they did. Some of my favorite posters at builder society (highly recommended reading if you still want to learn SEO) agree.
I won’t shut down my sites & don’t plan on selling (sites like mine trade for ~2.5x revenue or so, and mine aren’t quite big enough to be worth the hassle), but they’re in maintenance mode unless something changes drastically.
Please consider subscribing if you haven’t already. It’s more fun to write with an audience!