The SEO landscape changed underneath everyone. Here is what still works.

For twenty years, search engine optimization had a stable premise: someone types a query, gets ten blue links, and clicks one. Everything the industry built (keyword research, link acquisition, content calendars, rank tracking) was downstream of that premise.

The premise has changed. Not disappeared, changed. And the response we see most often from businesses is either denial ("SEO is fine, keep publishing") or panic ("SEO is dead, stop investing"). Both are wrong in ways that cost money.

Here is our read, from running search for clients through the shift.

What actually changed

The result page stopped being a list of destinations. AI overviews, featured snippets, People Also Ask, local packs, shopping units, and video carousels have pushed the first organic link progressively further down. For a large class of informational queries, the answer is now on the results page. The user got what they wanted and never left.

A second audience started reading your website. ChatGPT, Claude, Perplexity, Gemini and the AI features inside search now read pages and synthesize answers, citing some sources and ignoring others. These systems are not humans with a browser. They arrive without cookies, often without JavaScript execution, and they judge in seconds whether your page can be parsed, trusted, and quoted.

Volume stopped being the strategy. When the marginal cost of producing an article approached zero for everyone simultaneously, publishing more of it stopped being a differentiator. The web got a great deal more content and not much more insight, and ranking systems adjusted toward signals that are expensive to fake.

Intent got sorted. Informational queries increasingly resolve on the results page. Transactional and local queries still send clicks, because a person who wants a plumber in Boca Raton on Tuesday cannot be satisfied by a paragraph.

What did not change

This is the part the "SEO is dead" crowd gets wrong.

• Being the best answer still wins. The mechanism for delivering the answer changed. The reward for having it did not.
• Technical health still gates everything. A slow, broken, uncrawlable site loses in every system: traditional index, AI overview, and answer engine alike.
• Local intent still converts. Nobody asks a chatbot to unclog a drain.
• Links still function as trust. Fewer, better, genuinely earned, but the signal has not gone away.
• Commercial queries still send traffic. People buying things still click, compare, and read reviews.

What to actually do

1. Grade the site for machines, not just for people

Assume two visitors: a human on a phone, and a model with a few seconds and no patience. Both need clean HTML, real headings, content in the markup rather than assembled by client-side JavaScript, fast responses, and structured data that says what the page is.

We check this so routinely that we built a tool for it: our website grader scores speed, code health, SEO and AEO (answer engine optimization) separately, because a site can be excellent for Google and illegible to an answer engine.

2. Decide deliberately what you let AI crawlers do

robots.txt now governs more than Googlebot. GPTBot, ClaudeBot, PerplexityBot and others respect it. Blocking them all is a real choice with a real cost: you cannot be cited by a system you have excluded. Allowing them is also a choice. What you should not do is make it by accident, which is what most sites have done.

3. Publish things that require having done the work

The content that survives is the content a model cannot synthesize from the other nine articles on the topic: your data, your process, your outcomes, your specific numbers. Original research, real case studies, actual pricing, honest comparisons. Everything else is a commodity being produced at infinite scale.

4. Answer the question in the first hundred words

Both audiences benefit. The human gets what they came for; the model finds a clean, quotable statement rather than reconstructing your argument from an anecdote about your childhood.

5. Change what you measure

Rank tracking alone is now actively misleading. Position one on a query answered on the results page is a vanity metric. Track branded search volume, direct traffic, assisted conversions, and citations in AI answers, and hold organic accountable to pipeline, not to sessions.

If you cannot connect a search investment to revenue, you do not have an SEO problem. You have an attribution problem wearing an SEO costume.

6. Stop treating SEO as a separate channel

The technical work that makes you fast makes your ads convert. The content that earns citations feeds your email. The reviews that win the local pack raise close rate on every channel. Search performance is increasingly a symptom of overall business legibility rather than a discipline you buy in isolation.

The uncomfortable summary

Search is delivering less free traffic per unit of content than it used to, and that trend is not reversing. The companies doing well are not the ones publishing the most. They are the ones with something specific to say, a site that machines can read, and a clear line from search to revenue.

If you want to know where your site currently stands with both audiences, run it through our free grader: it takes about a minute and tells you what a model sees when it reads you. If you would rather have someone own the number, apply for partnership.