I’ve spent 11 years inside SEO. I’ve sold it, delivered it, defended it, built an agency around it and watched a small army of LinkedIn gurus slowly turn it into a content sausage machine with a keyword tool attached.
So let me be blunt. The SEO model most businesses still understand is dying.
Websites aren’t going anywhere, and content isn’t useless. The old bargain that made SEO commercially worth doing is breaking.
The bargain was simple. Create useful content. Google indexes it. People search. Your page ranks. You get traffic. Some converts. The agency sends a graph that goes up and to the right. The client pretends to understand the graph. The retainer renews.
That world isn’t gone yet. But it gets less reliable every month.
Google is no longer just a search engine. It’s an answer engine, a task engine and, increasingly, a recommendation engine. Google’s AI Mode hit one billion monthly users in May 2026, with more conversational and agentic features built directly into Search. Once the answer is inside the interface, the click becomes optional. When the click becomes optional, the whole SEO industry needs to stop pretending this is just another featured snippet problem with a slightly shinier hat.
The unanswered question
A few days ago I saw a question in Google’s own Search Console help forum that captures the problem cleanly.
If Google tells us to create original, helpful, non-commodity content, but then Google’s AI summarises that content and gives the answer away inside Google, where does the traffic come from?
It’s the right question. And the silence around it says everything.
Helpful content made sense when helpfulness was rewarded with traffic. You gave Google something useful. Google sent you people. There was an exchange. Now the exchange is changing. AI systems still need your knowledge – they need the raw material of human expertise, experience, examples, reviews, cases, opinions and proof. But they don’t necessarily need to send you the visitor.
That’s the bit too many SEO people are politely avoiding, because it makes the old pitch sound ridiculous.
“Create more content,” they say. Fine. For whom?
“Get cited in AI answers,” they say. Fine. To achieve what?
“Optimise for generative engines,” they say. Fine. But if the outcome isn’t traffic, what exactly are we selling?
A lot of the SEO influencer community is currently tying itself in knots trying to turn AI search into another excuse to sell blog content. More articles. More citations. More entity optimisation. More schema. More tactical window dressing that’s presented as a new discipline.
Some of that will matter. Technical visibility, clear pages and brand mentions still matter. But none of it is the main shift.
The main shift: in the old world, you tried to rank. In the new world, you need to become worth recommending.
Traffic is becoming a bonus
Seth Godin recently argued that creators should treat traffic as “a random gift, not a resource to be depended upon”, serve the smallest viable audience and earn permission to follow up directly. That’s not a cute creator economy observation. It’s a serious strategic warning.
For years, too many businesses built on rented distribution. Google was the landlord. So were LinkedIn, Instagram, TikTok, Facebook and YouTube. Useful, powerful and completely not yours.
Now the landlord is changing the locks again.
Pew Research found that when Google users encountered an AI summary, they clicked a traditional search result in 8% of visits, compared with 15% when no summary appeared. They clicked a source link inside the AI summary in just 1% of visits. The direction of travel is clear. More answers will get resolved inside interfaces. More buying decisions will involve someone asking ChatGPT, Gemini, Claude, Perplexity or Meta AI what they should choose.
The neat world of “rank position three and harvest the demand” looks increasingly fragile.
The future isn’t one search engine and ten blue links. It’s a messy network of AI recommenders. And the businesses that get recommended are the ones that have done something to deserve it.
Helpful content isn’t enough
This will annoy some people, so let’s be precise.
Helpful content isn’t dead. Boring helpful content is. Generic helpful content is. Low-effort “how to” content written because a keyword tool found 900 monthly searches is – even if it keeps twitching in the analytics for a while.
AI is already very good at basic helpfulness. It can tell you how to clean an air fryer, explain how to run a sales meeting, build a simple calculator, produce a checklist, a comparison table, a template, a cold email and a 12-step framework before your SEO executive has finished naming the Google Doc.
If your strategy is built around answering simple questions at scale, you have a structural problem.
The future isn’t content that helps someone know something. It’s content and assets that help someone transform something: their thinking, their performance, their results, their decisions.
People don’t need another bland answer. AI can supply bland answers all day. What they need is judgement. Experience. Proof. A point of view. A body of work that helps a specific group of people achieve a specific result.
That’s not content marketing in the old SEO sense. That’s utility.
Utility with effort in it
The word utility gets abused, so let’s make it practical.
Say you run a speakeasy cocktail bar in Manchester. The old SEO brain says: write “best cocktails in Manchester”, “date night bars Manchester”, “how to make an old fashioned” and 47 other posts that look exactly like every other bar’s posts.
What would actually be useful? Create a weekly email showing people one cocktail they can make at home this weekend. Film your bartender making it. Give people a shopping list. Run a monthly tasting night. Build a small community of people who care about cocktails, not just a database of people who once searched “nice bars near me.”
That’s still content. But it’s a useful asset tied directly to the business, not content for content’s sake.
Now imagine a meeting intelligence SaaS for sales teams. The lazy version is a blog called “10 ways to run better sales meetings.” You’ve created digital wallpaper.
The useful version runs free training on discovery calls, publishes anonymised patterns from successful meetings, runs monthly webinars with proper sales operators, hosts a sales event and generally helps salespeople convert more, specifically, not vaguely. Because the software records meetings, but the customer wants to win more revenue. The brand should become useful around that outcome.
Or take an air fryer brand. One recipe is nothing. Ten is not much better. But a hundred and fifty chef-led recipes, organised by occasion, diet, budget and skill level, with videos, emails and a proper reason to keep coming back – now you have an asset. You’re not selling a machine that makes chips with less guilt. You’re helping people turn boring beige meals into something worth making and sharing.
The rule behind all of this: if AI can make it in two minutes, it’s not a content asset. It’s digital litter.
Articles still matter. Bad articles don’t.
Some people will read this as “articles are dead.” They’re not.
Long-form writing can change how people think. A newsletter can build a direct relationship with a specific market. A blog series can become the place a niche audience goes to understand its own problems. A book can reposition a company, a founder or an entire category.
But that only happens when the writing has a clear audience, a clear point of view and a clear transformation it’s trying to deliver. You’re not writing for everyone. You’re writing for the people you can help better than anyone else.
That’s the bit most SEO content missed. It chased demand instead of serving an audience. It wrote for keywords instead of customers. It answered questions without building memory.
Machines are perfectly happy to eat those articles, summarise them and serve them back inside someone else’s interface. The question is whether the article does anything more than provide raw material. Does it make someone think differently? Does it give them language they repeat? Does it make your brand easier to understand?
A single article can be useful. A serious body of articles built over years for a specific audience with a specific problem can become a distribution channel, a trust engine and a genuine brand moat. That’s the difference.
Stop writing generic content for imaginary search traffic. Start building assets that help a specific audience achieve a specific transformation.
Visibility compounds
When you create something genuinely useful, it moves.
A cocktail night gets talked about. People share photos. The bartender becomes recognisable. The bar becomes associated with a particular kind of taste and crowd.
A sales event creates clips, posts, referrals, attendees and authority. A recipe platform creates repeat visits, social sharing and email habits. A serious newsletter creates language, memory, trust, replies, forwards and sales conversations.
Nothing happens unless something moves first. That sounds obvious, but most businesses want visibility without doing anything visible. They want recommendations without becoming recommendable.
Visibility comes from signals. Signals come from activity. Activity compounds when it’s consistent, distinctive and useful to the right audience. The visible become more visible.
This is where AI visibility becomes brand marketing. Not fluffy brand marketing. Not a mood board with a manifesto and no sales figures. Brand marketing with a hard commercial outcome: become known for the outcome you deliver.
The cocktail bar known for the best modern cocktails in Manchester. The meeting intelligence platform known for helping sales teams win more deals. The air fryer brand known for making everyday food genuinely better, not just faster.
Agencies have to stop selling channels
For years, agencies have sold channels because channels were easy to package. SEO. Paid search. Paid social. Organic social. Content. Digital PR. Email. Nice neat boxes. Nice neat retainers. Nice neat reports.
Clients don’t want channels. They want outcomes. Visibility. Demand. Preference. Recommendations. Sales.
The channel was always just the route. And when the route changes, selling the route becomes a precarious business.
If Google sends less traffic, clients won’t carry on paying for SEO activity just because an agency can produce ranking reports. They’ll ask harder questions. Are we more visible? Are more people searching for our brand? Are we showing up in AI recommendations?
Good. That’s where marketing should have been all along.
The agencies that survive will combine strategy, positioning, content, digital PR, creative production, useful assets, distribution and measurement around commercial outcomes.
Not “we publish four blogs a month.” The offer has to become: we’ll help you become more visible to humans and machines for the thing you want to be chosen for.
Harder to sell. Harder to deliver. Doesn’t fit neatly into a keyword dashboard. Requires better thinking, better creativity and more genuine expertise.
Which is exactly why it will be valuable. Easy work gets automated. Hard work gets paid for.
What should we measure?
This new world will scare clients because it’s less tidy. Rankings were tidy. Traffic was tidy. Even when the numbers were misleading, they looked reassuring in a report.
AI visibility is messier. That doesn’t mean it’s unmeasurable. You still look at rankings, organic traffic, conversions and pipeline. But you also need to watch different signals.
Brand search growth. Direct traffic. Email list growth. Quality of mentions. Share of voice. Reviews. Referral traffic. AI recommendation visibility. Repeat engagement with owned assets. Event attendance. Newsletter replies.
The question changes. It’s no longer only “did the page rank?” The better question is: are we becoming the obvious choice?
A recommendation isn’t a ranking position. It’s a judgement. And judgements are shaped by memory, trust, proof, clarity and repetition.
The new rule
The old SEO model was built around being found. The new model is built around being chosen.
That doesn’t mean technical SEO disappears. Articles don’t disappear. Websites don’t disappear. You don’t stop trying to earn visibility wherever people are searching.
The centre of gravity moves. From keywords to customers. From traffic to trust. From rankings to recommendations. From content volume to useful assets. From being findable to being known for something specific.
That last one matters most.
Known to the people who matter. Inside the buying situation. For the job you actually help someone get done.
Because when someone asks an AI what to buy, who to trust, which platform to use, which bar to visit, which agency to hire – you want to belong in the answer.
You don’t get there by publishing more average content.
You get there by building a body of useful, effortful, distinctive work that helps a specific audience achieve something they care about.
Stop trying to rank pages. Start becoming worth recommending.