H&M SEO lead urges retailers to prepare for AI shoppers

H&M’s acting global SEO lead, Ian Macfarlane, is urging retailers to prepare for AI systems that recommend products and could eventually buy on customers’ behalf, ahead of his appearance at CustomerX on 14 October.

In an interview with InternetRetailing, he argues that retailers must protect their visibility as technology platforms take a larger role between brands and shoppers. However, he cautions that consumers’ willingness to entrust payments to those systems remains uncertain.

What happened

Macfarlane identifies a change in the audience retailers need to reach: alongside persuading people, brands may increasingly have to influence software making decisions for them. He considers that prospect particularly challenging for established businesses with valuable customer relationships to protect.

He says generative engine optimisation, or GEO, should sit alongside conventional search engine optimisation and search within retailers’ own websites. Depending on where customers look for products, the remit could also cover social platforms, other retailers’ search tools and local results.

Existing SEO teams are well placed to carry out that work, in his assessment. Their experience of coordinating technical and commercial changes across departments remains relevant, although the people involved increasingly include public relations and wider brand reputation teams.

The operational task extends beyond identifying what improves visibility. Search specialists must explain the value to colleagues, establish which teams need to act and make sure the changes are delivered. Macfarlane does not see AI search as removing that coordinating role.

This broadening remit comes alongside changes elsewhere in digital marketing, including UK retail media spending reaching £2bn after a 32% rise. His central point is that search activity should follow the customer rather than remain confined to a single platform or format.

The background

Macfarlane places AI within a longer evolution of search. Retailers initially competed for positions in lists of website links, then adapted to shopping results, maps and answer boxes. The latest stage requires them to secure recommendations, not simply appear among possible destinations.

Each development has reduced brands’ direct control over their presentation, he says. Google’s AI Overviews are a particular reputational concern because searches about a business can generate summaries that set out both favourable and unfavourable assessments.

Buying agents would take that separation further. Macfarlane expects software acting for a shopper could respond more rationally, and less emotionally, than a person, changing the approach required from marketers. He presents that as a potential development rather than an established pattern of consumer purchasing.

For now, he regards research and product recommendations as more proven applications. AI tools can review product feedback and bring together opinions that a shopper would otherwise have to examine individually. He says ChatGPT has increasingly focused on recommendations, while completing purchases has proved more difficult.

His own experience illustrates the narrower use case. Earlier in the year, he gave an AI tool a set of requirements for a new phone and received a recommendation that substantially shortened his research. That involved assistance with selection, rather than permission for software to spend independently.

What people are saying

Macfarlane argues that the value of delegated research depends on the consequences of a poor choice. A shopper considering an inexpensive item may accept a recommendation instead of reading hundreds of reviews; someone choosing a car is more likely to investigate further.

That distinction leaves a role for human judgement when using AI outputs. He says consumers must weigh the risk of an incorrect answer, rather than assume that the convenience of a recommendation makes it sufficient for every purchase.

Brand recognition remains useful in that process. A familiar name can give a customer confidence that a recommendation belongs on their shortlist, while an unfamiliar supplier may prompt further checks. Macfarlane also identifies a strong, positive reputation as an important contributor to visibility in AI search.

Fashion presents additional considerations because preferences are personal. He says conversational systems can potentially draw on more individual context than traditional search, including previous purchases and information in emails, allowing more tailored responses.

That does not remove the purpose of shops. Customers may be comfortable ordering socks online but want to assess the fit or feel of another garment in person. The relevance of a store therefore depends on the customer, the circumstances and the product, rather than a single preference for online or offline shopping.

Distinctive stores can also encourage visits and attract social media or press attention, he says. However, he acknowledges that creating such experiences is easier for higher-margin brands with impressive flagship premises than for businesses selling more functional products.

His more immediate recommendation is to maintain accurate store details across the sources used by Google, ChatGPT and other large language models. Incorrect opening information can send shoppers to a closed branch, while missing information can leave a competitor as the suggested destination.

What happens next

Macfarlane will discuss AI-powered search and agentic commerce at CustomerX’s CommerceAI thinktank on 14 October. The session will examine how customers’ use of AI could change retailers’ relationships with them.

He advises businesses to explore integration requirements now so they are not left facing a development programme lasting more than three years if demand accelerates. That preparation does not depend on committing to a confident forecast: he considers a five-year prediction of AI marketing too unreliable to guide detailed planning.

Trust remains an obstacle to adoption. Macfarlane says AI’s capacity to produce conspicuously wrong answers makes it difficult to judge when people will feel comfortable delegating payments. The wider issue of software acting for users is also drawing ICO scrutiny of AI agents.

Another practical priority is understanding the occasions behind searches. He says AI can handle detailed requests about particular events or circumstances more effectively than traditional search, although many consumers still use chatbots much as they would a search engine.

Retailers should identify the situations that matter to their customers, establish where AI platforms obtain answers and then address those sources through relevant website content, public relations or other activity. Macfarlane’s recommendation is to base that work on specific customer needs, rather than assume everyone has reached the same level of familiarity with AI.