The original internet story was disintermediation. Producers would sell directly to customers, eliminating middlemen. But with abundant choice, how would you know where to look for what you wanted?

When I was a child, my mother would take me into the city centre once a year, towards the end of the summer holidays. We’d browse the shops selling similar products. Hours later, we’d struggle home laden with new school uniform, sports equipment and a fresh stock of stationery. When my own children needed these items, we’d order them over the internet.

Savvy retailers realised they needed a digital shopfront. The B2C consumer website was born. But the benefit of city centre shopping was that competing products were all more or less in one place. The internet required an online equivalent.

Amazon became the new gatekeeper for retail. Apple became the gatekeeper for mobile software. Google collected information and Facebook curated attention.

New industries sprang up around these gatekeepers. Experts explained how to game Google. Advertising firms specialised in search engine optimisation. Marketing became much more technical and technological, deploying data science and delving into matching algorithms.

Every generation is told that a new technology will remove intermediaries. In reality, technology tends to move the choke points.

A recent Pew Research study showed that only 16% of Americans expect AI to benefit society over the next 20 years. Yet 44% use ChatGPT, more than double the 2023 number.

A small business may ask “Which accountant should I hire?” or “What’s the best CRM for a 50-person engineering firm?”. On the surface this is a win. There’s no need to wade through adverts, SEO pages and sponsored content. The LLM provides a recommendation.

The obvious question is where this recommendation is coming from. Influencing AI search becomes the new pinch point.

Once more companies will turn to their websites. These now need to be optimised to be read by machines rather than humans. There will be an extended messy period when companies are trying to appeal to both. There will also be payment for preferential treatment. There should be lots of advisory work about how to do those two things well.

None of this means the best products will win. That honour will go to companies that are easiest for AI to understand and recommend. The big change from what happens today is that there may be fewer winners.

Amazon rewards products with clear reviews, strong sales histories and structured data. It relies on people knowing what to type in the search bar to get what they want. It’s good for buying but less so for browsing. Google meanwhile rewards pages that fit its understanding of authority and relevance.

AI systems will develop their own preferences. Distribution may become more concentrated, which will pose a problem for small business owners without marketing nous or a big budget.

In the world of B2B, buyers search up to ten sites before approaching a potential supplier. If they switch to LLM search they might be guided to three sites. Being on that list is going to become your greatest challenge.

Nobody yet knows how AI-driven discovery will work. For now, we are assuming a significant overlap with what search and recommendation engines have rewarded.

This means becoming machine-readable. Websites designed for humans often include statements such as “delivering innovative solutions” or “helping manufacturing businesses deliver growth”. AI does not know what that means.

Websites for machines may be more straightforward and cut to the chase. They will answer questions such as who you are, what you do, who you do it for, why you are different and what results you have generated. Providing a price may help.

AI appears to favour evidence over promotional language. This places a priority on case studies, testimonials and technical product descriptions. The key will be to guide humans to the narratives, while steering machines to the details.

A third step is to build your brand away from your website. AI search will use LinkedIn, industry publications, review sites and podcasts. Reputation will be determined by the number and variety of times that others talk about you.

The one thing to avoid is chasing cheap tricks. These do not work for long. Quality content about solving the problems of genuine customers will trump AI-specific keywords and formatting tricks.

This means that the companies best prepared for AI search will be those that are doing what they should have been doing all along. The technology may turn out to be a lot less scary than people fear.

When incentive structures change they key is not to be distracted by the potential downside. The winners will be those who identify where the new bottleneck will form.