Does AI Content Actually Work for SEO?

There is no shortage of opinions about AI-generated content and the platforms that will do it for you.

I’ll admit, we’ve had customers over the last few years ask us to simply generate new blog posts using AI.  So we’ve used a wide range of the top AI content generation platforms out there.  But at the end of the day, we didn’t see the SEO needle make any movement.

Don’t get me wrong, the content seemed to be good, but it just didn’t deliver.

Some marketers believe AI lets you scale SEO faster than ever. Others are convinced Google will eventually punish websites for using it. And plenty of companies have taken the simplest possible approach: produce as much content as possible and hope something sticks.

We wanted a better answer.

So instead of debating whether AI content works, we decided to test it. Over a period of roughly two months, we created 224 new AI-assisted pages across five very different websites:

1. A Weight Loss Clinic
2. A Cardiovascular Medical Practice
3. A General Medical Practice
4. A Startup SaaS Company (we are talking a brand new website here from scratch)
5. A Software Agency

These are not big sites, nor do they have huge traffic numbers.  We wanted to see what happened for the average website.

The sites serve different audiences, operate in different industries, and started with very different levels of search visibility.

That made the experiment much more useful. If one site improved, it could simply be a lucky topic or an existing strength of the domain. And importantly, we looked at the most recent 28 days compared with the preceding 28 days.

That last comparison matters because SEO does not move in a perfectly straight line. A longer before and after comparison can sometimes hide what is happening right now. If similar patterns appeared across several unrelated websites, we would have something much more interesting to study.

We pulled approximately 12 months of Google Search Console data for the five sites and compared what happened before and after the content generation programs began.

The results were encouraging.

But they were also much more complicated than simply saying, “AI content works.”

First, the headline numbers

Across the four websites, we published:

224 new pages

Those pages generated:

502,548 Google Search impressions

and

2,249 organic search clicks from Google

within the 2-month test period.

That alone is noteworthy.

Of the 224 pages:

125 are averaging a Google position of 10 or better.

189 are averaging a position of 20 or better.

In other words, about 56% of the new pages are already averaging a first page position, based on Google Search Console’s average position metric.

About 84% are averaging within the top 20.

That does not mean every page is a success.

Far from it.

Only 104 of the 224 pages had generated at least one click at the time of the analysis. That difference between rankings, impressions, and actual clicks became one of the most important things we learned.

The results were not the same on every site

Here is the basic picture (Strictly Google Organic Traffic).

Site	Clicks Before AI Program	Clicks After AI Program	Change in Clicks	Change in Impressions	Clicks Directly From New AI Pages	AI Pages as % of Post Period Clicks	Most Recent 28 Day Click Trend
Weight Loss	1,782	2,775	+55.7%	+31.1%	520	18.7%	+59%
Cardiologist	5,808	6,382	+9.9%	-20.3%*	1,382	21.7%	+67%
General Practice	1,023	1,564	+52.9%	+7.6%	225	14.4%	+63%
Click Laboratory	282	335	+18.8%	+73.3%	122	36.4%	+7%
Combined	8,895	11,056	+24.3%	—	2,249	20.3%	+62%

* The cardiologist’s impressions declined into June after an extremely strong February through April, but then reversed sharply. Its most recent 28 days showed impressions up about 57% and clicks up about 67% versus the preceding 28 days.

Across the four sites, organic clicks increased 24.3% after the AI content programs began. The 224 new pages directly generated 2,249 clicks, accounting for virtually all of the portfolio’s net increase in organic traffic during the comparable periods.

Recent performance is accelerating. In the most recent 28 days of the dataset, combined organic clicks across the four sites were approximately 62% higher than the preceding 28 days, while impressions were approximately 65% higher.

Across all four websites, organic clicks increased from:

8,895 clicks before the programs

to

11,056 clicks after the programs

That is an increase of:

24.3%

The 224 new pages themselves generated 2,249 clicks.

Here is where the math gets particularly interesting. If we subtract those 2,249 clicks from the post program total, the remaining portions of the websites generated approximately:

8,807 clicks

Before the programs began, those sites generated:

8,895 clicks

That is almost perfectly flat.

So mathematically, the traffic generated by the new pages accounts for virtually all of the net increase in organic clicks across the four sites. That does not mean every other page was unaffected by the content program. Some existing pages gained traffic, while others lost traffic. But it does tell us something important.

The growth was not simply coming from websites that were already growing. The new content represented a substantial new source of organic traffic.

The most recent data may be even more interesting

The before and after comparison gives us a useful long view. But we also wanted to know what is happening right now.

So we compared the most recent 28 days in the dataset with the previous 28 days.

Across all four websites combined:

Organic clicks increased approximately 62%.

Google Search impressions increased approximately 65%.

That recent acceleration appears on all four sites, although the magnitude varies considerably. This matters because some of the sites declined during the early part of the content program before beginning to move upward.

The Cardiologist is the best example.

Rankings are not the same thing as SEO success

One of the biggest lessons from this experiment is that traditional SEO metrics can easily mislead you.

A page can have:

• Thousands of impressions.
• An average first page position.
• Dozens of ranking queries.
• And almost no traffic.

That sounds strange until you consider how modern Google results work.

• A user may receive an answer directly from an AI Overview.
• The search result might contain maps, videos, images, featured results, Reddit threads, or other elements that receive the click first.
• Google may be testing your page for searches that are only loosely related to what the user actually wants.
• Your title may technically answer the query but fail to give someone a reason to click.
• Or the query itself may simply have low click potential.

That is why impressions and average position cannot be the end of SEO reporting.

We need to ask what happened next.

1. Did someone click?
2. Did they engage with the page?
3. Did they continue through the website?
4. Did they convert?

There is another reason average position can be misleading

Adding a large amount of content can actually make a site’s overall average position look worse while the site becomes healthier.

Why?

Because new content introduces the website to thousands of new queries. A website that previously ranked for 1,000 searches may suddenly start appearing for thousands more. Some of those new rankings might begin at position 15, 30, or 50. That can pull down the website’s average position while simultaneously increasing visibility and clicks.

For that reason, we care far more about the behavior of individual pages and query groups than one sitewide average ranking number.

What did all four websites have in common?

The individual stories were different. But several patterns appeared repeatedly.

1. The strongest content answered specific questions

Broad topics were not the obvious winners.

Specific questions were.

Not:

“Heart Health.”

But:

Why is my heart rate high and I am sweating at rest?

Not:

“Weight Loss Tips.”

But:

Can you notice a 10 pound weight loss?

Not:

“Acne Treatment.”

But:

Does sleeping on your side cause acne?

Specificity gives a page a clearer job.

It also tends to align much more closely with how real people search.

2. Expertise still matters

A Cardiologist can credibly answer cardiovascular questions.

A Weight Loss Clinic can credibly discuss weight loss.

A General Medical Practice can credibly answer patient health questions.

AI makes creating the content easier. It does not give a website authority it does not have. That is an important distinction.

3. A minority of pages generate most of the traffic

This happened on every website. Some pages became winners very quickly. Others ranked but did not get clicks. Some received impressions but never broke through. Some have not done much of anything yet. That is exactly why publishing cannot be the final step.

The performance data should determine what happens next.

4. Current momentum matters

A simple before and after comparison can miss what is happening inside the period.

The Cardiologist is the perfect example. Its full post program period still contains a substantial decline into June. But the most recent 28 days show clicks up 67% and impressions up 57%. Across the entire four-site portfolio, the latest 28 days produced approximately:

62% more clicks

and

65% more impressions

than the preceding 28 days. That tells us the content program is not simply living off a handful of pages that ranked early.

The sites, particularly the Cardiologist, Weight Loss Clinic, and General Medical Practice, are currently gaining significant momentum.

So, does AI content work for SEO?

Based on this experiment, the answer is:

Yes, but that is probably the wrong question.

AI is not an SEO strategy. AI is a production tool.

If you use it to create generic content nobody needs, you can produce bad content much faster than you could before.

That is not a competitive advantage.

If you use AI to help identify and answer hundreds of highly specific questions that are relevant to an organization’s actual expertise, the economics become much more interesting. That is what we are seeing. Across four websites, 224 new pages generated:

More than 500,000 Google impressions

More than 2,200 organic clicks

125 average first page rankings

and

189 average top 20 rankings

At the site level, comparable periods after the programs began generated approximately:

24% more organic clicks

than the periods immediately before the programs began.

And the most recent 28 days show even stronger momentum. That is enough evidence for us to continue. But it is not enough evidence to simply produce another 1,000 pages without thinking.

The worst thing we could do next is blindly scale production

It would be easy to look at 500,000 impressions and say:

“If 224 pages did that, let’s publish 2,000.”

We think that would miss the biggest lesson in the data. The first 224 pages gave us information.

Now we need to use it.

• We know which topics are getting impressions.
• We know which pages are reaching the first page.
• We know which pages generate clicks.
• We know which subject areas are producing disproportionate results.
• We also know which pages are being shown thousands of times without convincing anyone to visit.

That gives us a much smarter roadmap.

Create more content around proven winners

• If several symptom-based cardiovascular topics are performing well for the Cardiologist, that is a signal.
• If acne and sleep-related questions work for the General Medical Practice, that is a signal.
• If behavioral weight loss questions outperform generic dieting content for the Weight Loss Clinic, that is a signal.

Follow the data.

Improve pages with rankings but weak click-through rates

This may be one of our biggest opportunities.

A page already averaging position six with thousands of impressions has cleared a difficult hurdle.

Google believes it is relevant.

If nobody clicks, we should investigate the query mix, title, search result presentation, and intent before abandoning the page.

Improve pages sitting just outside the traffic zone

A page averaging position 11, 12, or 15 may be worth more effort than another brand new article. It already has visibility. A stronger answer, improved internal linking, better examples, useful visuals, better structure, or additional expert input could potentially move it into a much more valuable position.

Stop treating every page equally

Some topics simply have more potential. Our own data is now helping us find them. We should be willing to invest more in the winners and less in subject areas that repeatedly fail.

Connect SEO data to business results

This is the most important next step.

Google Search Console tells us who saw a result and who clicked. It does not tell us enough about what happened afterward. Now we need to connect these pages to GA4 and conversion data.

The next experiment is about revenue, not rankings

The question we are asking now is not:

Can AI content rank?

We already know it can.

One hundred twenty five of these pages are averaging positions within Google’s top 10.

The better question is:

What happens after someone clicks?

• We want to know which articles generate engaged sessions.
• Which readers visit service pages?
• Which articles contribute to appointment requests?
• Which pages generate phone calls?
• Which visitors join an email list?
• Which business articles contribute to leads?
• Which content actually produces customers?

That is the point where an SEO program becomes much more valuable. Because an impression is not revenue. A ranking is not revenue. A click is not revenue. They are steps along the way.

This is why we believe content needs to be alive

Traditionally, companies have treated content like a project.

Pick a topic.

Write the article.

Approve it.

Publish it.

Move on.

That model makes less sense every year. AI makes creating and updating content faster. Analytics give us more information about how people interact with it. Google continuously changes. Competitors continuously change. Search behavior continuously changes.

So the content should change too.

We call this living content.

The idea is simple. Publish the best page you can. Watch what happens.

• If Google begins showing it, learn which searches are generating those impressions.
• If it ranks but does not get clicked, figure out why.
• If users click but leave, improve the page.
• If the page begins producing conversions, expand around the topic.
• If performance starts declining six months later, update it.

The article is not finished because someone pressed Publish.

Publishing is where the data starts.

What we believe the data says so far

We are still early in this experiment. Some of these pages have only been live for a few weeks. SEO often takes months to mature. That means we are not ready to declare every question answered. But there are several things we feel comfortable saying.

AI assisted content can rank. We have 125 new pages averaging positions within Google’s top 10.

AI assisted content can generate meaningful organic traffic. The 224 pages have already produced 2,249 clicks.

The results vary dramatically by topic and website. Publishing volume alone does not guarantee success.

Specific questions appear to outperform generic subject matter. This pattern appears repeatedly across the strongest pages.

Rankings without clicks are not enough. The Software Agency demonstrates that better than any of the other sites.

The current performance trend is positive across all four websites.

Every site generated more impressions during the most recent 28 days than during the previous 28 days.

The Cardiologist, Weight Loss Clinic, and General Medical Practice also generated dramatically more clicks.

And most importantly:

AI appears to be most valuable when it makes the entire content learning cycle faster, not simply the writing process.

Creating 224 pages gave us 224 opportunities to learn.

We now know more about what people search for.

We know more about what Google rewards.

We know more about what gets clicked.

And we know where the gaps are.

That information should make the next 224 pages much better than the first 224.

The real competitive advantage is not AI

Every company is going to have access to AI.

That means simply using AI will not be much of an advantage for long.

The advantage will come from what companies build around it.

• Better research.
• Better subject matter expertise.
• Better topic selection.
• Better analytics.
• Better conversion tracking.
• Better testing.
• And a willingness to continually improve content after it goes live.

AI lets us do more.  The decision to use it or not is up to each individual company.

Data tells us what deserves more.

The combination is where things get interesting.

Our first 224 pages have generated more than half a million Google impressions and over 2,200 clicks.

Across the four sites, organic clicks increased 24% in comparable before-and-after periods.

And in the most recent 28 days, clicks across the portfolio increased another 62% compared with the previous 28 days.

Those results give us a reason to keep going.

But the goal is not to see how many AI articles we can publish. The goal is to find out whether a smarter content system can continuously create more search visibility, more qualified traffic, and ultimately more business.

That is the experiment we are really running.