By Andy Squire, RVP EMEA, Brave
Search used to have a visible interface. A person typed a query, scanned results, clicked a link, and left a trail to measure. That picture is starting to feel too small, because the next searcher may be a human, a summary engine, or an AI agent.
It might sound simple, but if marketers are to stay ahead, they need to understand how the LLMs powering these answers work, and how to optimise for them.
Look at the index instead of the search results
Most search teams are still doing useful work inside familiar systems. They’re managing bids, refining copy, protecting ROAS, and finding efficiency in crowded auctions. Nobody should dismiss that work, because it still captures demand, and it still pays.
The trouble is that search is becoming less dependent on the results page. A user may ask an assistant to compare products, plan a trip, shortlist providers, or explain which option looks credible. The answer may be a paragraph, a recommendation, a shopping flow, or a prompt to take the next step.
That’s why the search engine index is becoming more important than the interface. LLMs didn’t become useful for current decisions by magic, and as we all know by now, they can’t rely only on training data. They need fresh, structured access to the web, so search starts to look less like a media channel and more like infrastructure.
Prepare for the invisible robo-shopper
Early AI tools are already starting to fill forms, book services, create carts, and help users buy. The consumer is still the decision-maker, but the route between intent and action is becoming more automated, which changes the basic ingredients of search marketing. A landing page written for a human reader may still be essential, but it won’t be enough on its own. Product feeds, availability, pricing, reviews, policies, and fulfilment details need to be clean enough for agents to interpret.
This is where many brands will get caught out. They’ve spent years polishing the customer-facing website, while the data routes used by comparison engines, answer engines and agents are incomplete, messy, or out of date. If an AI agent can’t read the product, confirm the stock, understand the return policy, or complete the next step, it may never recommend the brand.
Follow the intent, not the interface
The rise of AI answers doesn’t mean every AI surface deserves a budget tomorrow morning. Some answers will be research-led, some will be informational, and some will carry clear commercial intent. The job is to understand where that intent appears, how it’s grounded, and whether the brand is represented properly.
As marketers we know that search advertising works best when it responds to clear intent. If someone is researching the history of hiking boots, that’s one kind of signal. But if they’re asking which boots are available in their size, for delivery this week, that’s a very different signal.
Live inventory is a good example. While organic content can feature details about an event, a hotel, or a product category, it likely won’t know whether seats, rooms, or stock are available now. In those moments, fresh commercial data can make the answer far more useful, because the user suddenly has something they can act on.
Give independent search the resources it deserves
Search infrastructure is concentrated. In the past, that might have felt convenient, because the big platforms offered scale, familiar tools, and defensible reporting. However, convenience can turn into dependency when search becomes the data layer behind assistants, browsers, answer engines, and agents.
Independent indexes are really important because they give machines more than one route into current web knowledge. They widen the information available to AI builders, commerce platforms, and advertisers, without routing every decision through the same dominant systems. That can affect resilience, source diversity, competition, and the commercial routes through which intent is discovered.
The good news is that independent search infrastructure like our own now exists on a commercially relevant scale, with large web indexes, billions of monthly queries, and LLM-focused APIs. And privacy-preserving search can still be useful when relevance comes from present intent, not third-party cookies or behavioural profiles.
The issues that leave a bad taste in traditional search will inevitably be present in agentic commerce search as well. Dominant search players rely on user tracking and profiling to power ads and personalise results. The market deserves players who, like Brave, don’t collect personal data or build user profiles. Independence means choice: allowing users to be safer online, and not be beholden to the privacy invasions, censorship, biases or economic interests of Big Tech.
Measure the market you’re missing
While we shouldn’t abandon existing search activity altogether – the largest platforms still deliver enormous volume and proven returns – the best approach is to stop treating those platforms as the whole map. Instead, ask where your data is visible outside the obvious dashboards. Explore whether alternative search environments add demand you’re not already reaching – and also pressure-test whether your commerce infrastructure can serve an agent as well as it serves a person.
As this new era of search begins to dawn, there’s a measurement problem coming too. Human search tends to leave familiar signals, including impressions, clicks, sessions, baskets, and conversions. Agentic search may involve multiple retrieval and verification steps before one recommendation appears, so marketers will need to think harder about coverage, feed quality, answer inclusion, and the data behind decisions.
Search is becoming bigger than the box where people type questions; increasingly, it’s the infrastructure that helps machines retrieve, compare, recommend and act. And it’s time brands start to harness the growing power of LLMs. Those who act early will make themselves easy to understand, easy to verify, and easy to choose – whether by a human or a fellow agent.
