SEO in 2027: What Is Actually Changing

Most "SEO trends" pieces are a list of things that were already true last

year. This one is limited to changes we can actually point at, with the

measurements that support them, and it says plainly where nobody knows yet.

1. Being indexed stopped being the finish line

The measurable shift of 2026 is that ranking and being cited came apart.

We tested this on ourselves. Our own site is technically clean: every AI

crawler gets a proper server-rendered page, the structured data is in place,

nothing is blocked. Ask Perplexity the exact question our product exists to

answer and we are not mentioned. The answer is assembled from other people's

articles.

That gap is the thing to plan around for 2027. Technical readiness is a

floor, not a score. It gets you eligible and does nothing to get you chosen.

What to do about it: stop treating AI visibility as a technical project.

The work that changes it is being the source other people cite, which is a

content and PR problem wearing an SEO hat.

2. The measurement problem is not solved, and will not be soon

Google does not expose AI Overview appearances through the Search Console

API. It is available in the Performance report under the Generative AI

filter, as an export only.

So through 2027, any product claiming a fully automated AI Overviews feed is

either scraping results, which measures a simulated search rather than

yours, or reporting something adjacent and calling it that.

What to do about it: ask any vendor where the number comes from, and

treat a vague answer as an answer. We import ours from your own Search

Console export, which is one manual step and your actual data.

3. Crawlers stopped behaving alike

We fetched our own site as four different AI crawlers and logged every

response. They do not behave like Googlebot and they do not agree with each

other. Some do not execute JavaScript at all, so a React or Vue application

returns close to nothing to them.

Googlebot has rendered JavaScript for years, which lulled a lot of teams into

thinking the problem was solved. For AI crawlers it is not.

What to do about it: if your site is JavaScript-rendered, check what each

crawler actually receives rather than assuming. Server-rendered HTML is worth

more in 2027 than it has been for a decade.

4. Click-through rates keep falling, and position means less

More of the result page is taken up by things that are not organic results.

Position three below an AI Overview, four ads and a video carousel is not

position three in any useful sense.

Our own numbers make the point: 38,100 impressions and 39 clicks over 28

days. That reads as a catastrophe until you notice the average position is

43, where almost nobody clicks anything. The click-through rate was never the

problem, and reading one number without the others is how people reach

exactly the wrong conclusion.

What to do about it: stop reporting position on its own. Impressions,

position and click-through rate only mean something together.

5. Where we genuinely do not know

Worth saying, because most trend pieces will not.

Whether AI search reduces total search volume. The confident answers in

both directions are guesses. The data is not public and the behaviour is

three years old.

Whether citation in AI answers converts. Being named in a ChatGPT answer

feels valuable. Whether it produces customers at a rate worth optimising for

is not something anyone has published credible numbers on.

What Google does with AI Overviews next. They have been added to and

removed from queries constantly through 2026. Building a strategy on their

current behaviour is building on sand.

What this means for a small team in 2027

The unglamorous answer is that the fundamentals got more valuable, not less.

Server-rendered HTML matters more because AI crawlers are less capable than

Googlebot. Clear answers under question-shaped headings matter more because

that is what gets extracted. Being cited elsewhere matters more because that

is what AI answers are built from.

None of that is new. What changed is that the cost of skipping it went up.