SEO Search Intent vs AI User Intent: Why “Know, Go, Do” Is No Longer Enough

“Set and Forget” AI? That Could Be Your First Mistake

For years, marketers have treated search intent like gravity.

Stable. Predictable. Always there.

People wanted to find things online, so we built websites around that behaviour. They wanted answers, so we wrote blogs. They wanted brands, so we built landing pages. They wanted to buy, book, download, or request a quote, so we added forms and buttons.

That model worked because the search engine was the front door.

But generative AI has changed the door …. And when the door changes, the shape of the room changes too.

If you think search intent does not change, or that people still have the same needs and only the starting point has moved, it is worth taking another look. The shift from search engines to AI chatbots is not just a channel shift. It is a behaviour shift.

Search engines helped people find.

Generative AI helps people make.

That difference is not cosmetic. It changes what users expect, what websites are for, and how businesses should design their digital presence.

We used to browse. Then we searched. Now we ask AI to do the job with us.

For the average person, online discovery began with directories.

Yahoo and other early web directories were like filing cabinets for the internet. Websites were sorted into categories. You clicked through folders of links and hoped the right thing was inside.

Then came search engines: WebCrawler, Lycos, AltaVista, Google, Bing.

Suddenly, the filing cabinet became a magic box. You typed in a few words and the search engine returned a ranked list of pages. Over time, search became better at understanding messy human language, not just exact keyword matches. Voice search pushed this further. Siri, Alexa, and Google Assistant trained people to talk to machines in full questions.

Then generative AI arrived.

Now people do not only ask where something is. They ask for a draft, a plan, a comparison, a code snippet, an image edit, a product recommendation, a contract summary, a campaign idea, a workflow, a joke, a table, a diagnosis of a spreadsheet, answers to homework questions, a birthday card message or a better version of something they already have.

The old web gave us links.

Search gave us answers.

AI gives us outputs.

That is why AI user intent is different from SEO search intent.

The old model: Know, Go, Do

Why AI user intent is not just search intent in a new outfit

A search engine returns options.

An AI assistant returns a working answer.

That answer might include a summary, a recommendation, a draft, a design, a table, a checklist, a plan, a piece of code, or a next step. In many cases, the user does not need to click through to ten websites because the AI has already compressed the research layer.

This creates a problem for websites built around the old SEO model.

Most websites were designed to catch informational searches. They were packed with “How to…” articles, introductory guides, keyword clusters, and blog posts written to win the top of the funnel.

That content had a job. It brought people in.

But in AI discovery, generic ‘Know’ content is often absorbed into the answer layer. The user asks the AI, gets the summary, and moves on.

The website may still be useful as a source. It may still shape the AI’s understanding. It may still be cited, retrieved, or used as background context.

But it may no longer receive the same visit.

That is the key commercial issue.

Your content may still influence the journey, while your analytics show less traffic.

So the old question, “How many people visited this blog?” becomes less useful.

The better question is:

Did this content make our business easier for humans and AI systems to understand, trust, and use?

The new intent: Act

Generative AI introduces a new kind of user intent.

Act

Act means the user wants to create, change, produce, generate, automate, remix, or complete something.

Examples:

“Code a simple app.”
“Create a meme video.”
“Change this photograph.”
“Make this funnier.”
“Turn this meeting transcript into tasks.”
“Draft a proposal for this client.”
“Build a workflow that follows up overdue invoices.”
“Rewrite this service page for a younger audience.”
“Create a campaign plan for a B2B SaaS launch.”
“Compare these vendors and recommend one.”

Act is not the same as Do.

Do intent is usually about taking a known action. Buy this. Book that. Download this. Register there.

Act intent is generative. The user wants an output that did not exist before.

That output may be text, code, an image, a plan, a workflow, a summary, a decision, a design, or a system action.

This is only possible because generative AI can work with the user during discovery. It can answer and produce in the same interaction.

That is why Act belongs in the intent model.

The early numbers: generative intent is already huge

Profound’s 2025 study of more than 50 million ChatGPT prompts found that generative intent was the largest AI search intent category at 37.5%. Profound describes this as users asking directly for concrete outputs: create X, draft Y, do Z.

Profound later grouped conversation intent into four core types: Commercial, Informational, Generative, and Conversational/Other.

That matters because it shows AI behaviour is not just a mirror of Google behaviour.

Traditional search was weighted toward information.

AI usage is weighted much more toward production.

In other words:

SEO was built around the person who asks, “What should I know?”

AI is being shaped by the person who asks, “Can you help me make this?”

That is a very different user.

And it needs a different website.

From 80/10/10 to 40/10/20/30

The classic SEO search intent model looked like this:

80% Know
10% Go
10% Do

The emerging AI user intent model may look closer to this:

40% Know
10% Go
20% Do
30% Act

This is not a fixed law. It is a practical working model.

It says: stop building websites as if users mostly want general information.

They still need information, but they also need action paths, decision support, and production tools.

Let’s break it down.

40% Know: fewer articles, stronger knowledge

‘Know’ content does not disappear.

People still ask questions. AI systems still need sources. Search engines still matter. Buyers still need to learn before they trust.

But the ‘Know’ layer needs to become more valuable.

Good ‘Know’ content should do more than explain terms. It should show how your business thinks.

The lazy version of ‘Know’ content is the generic blog post:

• “Top 10 winter jackets”
• “What is a capsule wardrobe?”
• “Benefits of protein powder”
• “How to decorate your living room”
• “Best toys for kids”
• “How to care for leather shoes”
• “Why buy Australian made?”
• “Benefits of shopping local”

These can still help, but only if they bring something specific to the table.

The better version is:

• “Which winter jacket should you buy for Melbourne rain, alpine trips, school pickup, or daily commuting?”
• “How to build a capsule wardrobe from 12 pieces already in your cart”
• “Whey, plant, collagen, or casein: which protein powder matches your training goal and dietary needs?”
• “How to choose living room decor based on room size, natural light, wall colour, and existing furniture”
• “Age-by-age toy guide: what to buy for toddlers, preschoolers, and primary school kids without wasting money”
• “How to clean leather, suede, and vegan shoes without ruining the material”
• “What ‘Australian made’ actually means — and how to check before you buy”
• “How AI shopping assistants decide which products to recommend, and why your product data matters”

This is ‘Know’ content that earns its keep because it reduces doubt, prevents returns, improves AI product recommendations, and gets the customer closer to buying the right thing.

Generic content teaches the category … Useful content teaches the buyer how to choose.

10% Go: brand navigation gets smaller, but sharper

‘Go’ intent does not need to dominate the site, but it needs to be brutally clear.

When someone searches for a specific brand, or asks an AI system about a specific company, the source material must be easy to understand.

A human or AI assistant should be able to answer these questions quickly:

• Who is your business?
• What does it do?
• Who does it help?
• What services does it provide?
• What makes your business different from a normal business?
• What industries or use cases fit best?
• How does someone start a project with you?
• Where is the proof?

Many businesses make this harder than it needs to be. They hide the answer under vague language. They say things like “we are a leader in our industry” or “we’re committed to providing high-quality products” and expect the reader to fill in the gaps.

That works badly with humans. It works even worse with AI.

Clear language is now a technical asset. If your website cannot state what you do in plain terms, do not expect AI systems to do the interpreting for you.

20% Do: users still need clean action paths

‘Do’ intent becomes more important in the AI era because users expect less friction.

Once an AI assistant has helped someone research, compare, and decide, they do not want to hunt for the next step.

They want the action to be obvious.

For a service business, ‘Do’ pages and flows might include:

• Book a strategy session
• Request a systems audit
• Get a workflow automation review
• Submit a project brief
• Download an AI readiness checklist
• Start a proof-of-concept
• Request pricing
• Send existing process documents
• Ask for a data model review

This is not just about adding more buttons.

It is about designing the next step around the user’s state of mind.

A user who has just asked an AI assistant, “Which company can help me automate our customer onboarding workflow?” does not want to read twelve more blogs.

They want to know:

• Can you do this?
• Have you done it before?
• What happens next?
• How much effort is needed from us?
• How do we start safely?

‘Do’ pages should answer those questions and remove drag.

30% Act: the new layer most websites are missing

Act is where the biggest change sits.

This is the part of the website that helps users make something, test something, generate something, or move a workflow forward.

These are not blog posts.

They are useful tools.

A blog says, “Here is what what’s important to us that you should know.”

An Act experience says, “Show us your what you need and we will help you get that done.”

That is the difference.

In the AI era, the best websites will feel less like magazines and more like workshops.

Why old Know content can become clutter

A lot of SEO content was built for a world where search engines needed pages to return.

That gave us a flood of articles written less for buyers and more for ranking.

Many websites now have hundreds of posts that do not support the brand, do not help buyers choose, do not connect to action, and do not add anything that AI cannot summarise in five seconds.

That kind of content becomes clutter.

It gets in the way of the pages that matter.

It can also make the business harder to understand. If your website contains 200 generic articles and only six clear service pages, both people and AI systems may struggle to work out what you actually do.

This is why website reengineering matters.

Not a redesign. A redesign changes the paint.

Reengineering changes the machine.

It means asking:

• Which pages help buyers choose us?
• Which pages help AI systems understand us?
• Which pages support a real action?
• Which pages show proof?
• Which pages are generic filler?
• Which pages should be merged, rewritten, structured, or removed?
• Which pages should become tools instead of articles?

The goal is not less content.

The goal is less dead weight.

AI is also changing Go intent

Generative AI does not only affect informational content; it is also affecting Go intent.

In classic search, if someone wanted to visit a brand or source, the search engine usually sent them there. In AI chat, the system may bring the content into the answer directly. The user may ask about a company, product, policy, or service, and the assistant may summarise the source in the chat.

That can be useful for users.

It can also reduce site visits.

This is one reason old traffic benchmarks are becoming less reliable. A brand may still be part of the answer, but the click may not happen.

The website becomes more like a data source, proof layer, and action endpoint than a simple destination.

This is also where law and policy may reshape the web again.

The EU AI Act includes transparency and copyright-related rules for general-purpose AI models, and the European Commission has worked on a Code of Practice for general-purpose AI providers.

In Australia, the government has been developing its national AI policy direction, including responsible AI use in government and broader AI adoption settings.

In the United States, AI regulation is developing through a mix of federal debate, copyright policy work, and state-level laws. The U.S. Copyright Office has been publishing its report series on copyright and AI, while state AI legislation continues to move across many categories.

The details are still moving, and each jurisdiction is taking a different path. But the direction is clear enough: training data, copyright, privacy, AI use disclosure, licensing, transparency, and compensation frameworks are all under pressure.

Once laws become clearer and enforcement grows, website traffic patterns may shift again.

Some content may be used less freely. Some sources may be licensed. Some platforms may surface attribution in different ways. Some businesses may choose to block, meter, structure, or commercialise access to their content.

So any strategy based on “traffic will behave like it did in 2019” is fragile.

The new job of the website

The website is no longer just a place people visit after searching.

It now has to serve four types of users:

1. Humans who visit directly
2. Search engines that crawl and rank
3. AI systems that retrieve, summarise, and recommend
4. Agents that may take action on behalf of users

Most websites were built for the first two.

AI user intent needs all four.

That means the website needs:

• Clear service definitions
• Strong brand language
• Structured data
• Use-case pages
• Proof pages
• Decision support
• Fast next steps
• Tool-like experiences
• Machine-readable meaning
• Human-readable explanations
• Safe, useful action paths

This is where structured data and interoperability become more than technical nice-to-haves.

They become marketing infrastructure.

If your website is a messy cupboard, AI will not magically turn it into a dazzling showroom.

If your website is a well-labelled system, AI has a better chance of understanding what belongs where, what your business does, and when to recommend you.

What Semantia recommends that businesses do now

Start with an intent audit.

Take every key page on your website and mark it as one of four types:

Know — teaches, explains, compares, or clarifies
Go — helps users find, trust, or understand the brand
Do — helps users complete a transaction or next step
Act — helps users create, generate, test, automate, or produce

Then check the ratio.

Many websites will find they are still close to the old 80/10/10 pattern. Too much ‘Know’. Too little ‘Do’. Almost no ‘Act’.

That is the warning sign.

Next, clean the ‘Know’ layer.

Keep the pages that have original value. Merge or remove weak posts. Turn broad explainers into decision guides, use-case pages, technical explainers, comparison pages, and proof-backed content.

Then strengthen ‘Go’.

Make the brand easier to understand. Use plain language. Explain the business clearly. Show who you help, what you build, what problems you solve, what makes your approach different, and how people start.

Then improve ‘Do’.

Make the next action obvious. Replace vague “contact us” pages with specific paths. A buyer should not need to guess whether they need a consultation, audit, proposal, pilot, or demo.

Finally, build ‘Act’.

This is the new work.

Create tools, agents, templates, calculators, upload flows, guided assessments, structured intake forms, and workflow builders.

Do not just tell people what is possible.

Help them make the first useful thing.

The practical model: Know, Go, Do, Act

The old search intent model was built for a web of pages.

The new AI user intent model is built for a web of outcomes.

That is why the formula is changing:

Know: 40%
Help users and AI systems understand the problem, category, method, and decision points.

Go: 10%
Make the brand easy to find, trust, and explain.

Do: 20%
Make transactions, enquiries, bookings, downloads, and project starts easier.

Act: 30%
Help users create, test, generate, automate, or move work forward.

This is the shift from content as reading material to content as working material.

And it is the reason every serious business should reassess its website now.

Not because SEO is dead … because search intent is no longer the whole story.

• People still want to Know.
• They still want to Go.
• They still want to Do.

But now, more than ever, they also want to Act.

The businesses that win will be the ones that stop treating AI as a new search box and start treating it as a new user behaviour.

Because in the AI era, discovery is not just about finding the answer.

It is about making the next thing happen.

What to do next …

Talk to Semantia about a $500 personalised AI User Intent Website Audit and discover how to realign it for higher AI understandability and recommendation.

We’ll even recommend suggestions for Act intent that your website visitors and AI will love.

Our self-service page for buying an AI User Intent Website Audit is still under construction …. but you don’t have to wait.

Get started now … call 1300 766 328.