September 28, 2026 · 8 mins read

Surviving AI search: five lessons from the Technology for Marketing stage

Every marketer has heard the claim that nobody uses Google anymore. At Technology for Marketing in London, three people who spend their days on the problem took it apart, and what they put in its place was more useful than panic.

James Hall, Strategic Director for Go-to-Market Strategy at WP Engine, hosted a panel with James Bavington, Chief Experience Officer at StrategiQ, and Chris Pitt, CEO at GAIN Performance. The brief was to make sense of AI search, the European Accessibility Act, and how we can be best prepared for this much talked about topic. Here's what we took away, with some additional data behind the points they raised.

1. AI search is a layer on top of organic search, not a replacement (yet)

James opened by describing a mixed picture. Informational sites and news publishers are hurting badly. Brand and e-commerce sites mostly aren't.

The numbers support him. Chartbeat data published by the Reuters Institute shows Google organic search traffic to more than 2,500 news sites fell by a third globally between November 2024 and November 2025, and by 38% in the US, though the report notes it isn't clear how much of that is down to AI Overviews. Across the wider web the drop is far smaller. Similarweb data published by Graphite found organic Google traffic to the top 40,000 US websites fell 2.5% in 2025 compared with the previous year.

That doesn't mean nothing is changing. According to Similarweb, AI Overviews went from appearing in 15% of Google searches to 43% in a single year. Google is also steering users firmly towards its conversational AI Mode. At I/O in May it made Gemini 3.5 Flash the default model in AI Mode globally and introduced the biggest redesign of the search box in over 25 years.

Chris argued this is the latest step in a long evolution rather than a break with the past. Search went from blue links to maps, then knowledge panels, then "people also ask", and now AI answers. LLMs still account for a relatively small share of search, though estimates vary widely depending on who's measuring. His advice was the least fashionable and most useful of the session: don't rush into a new tactic. Be decisive and specific instead.

GAIN's take: Don't panic, but don't wait either. Find out how much of your traffic is informational. That's your exposure, and it's where to start.

2. Accessibility and AI visibility are one strategy, not two

With the European Accessibility Act now in force, many marketing teams see accessibility as a compliance cost competing for budget with AI optimisation. Chris argued they're the same investment. You shouldn't design for Google, ChatGPT or Meta. You should design for your users, and machines are now among them.

The technical reason is that most AI systems read your site the way a screen reader does. Vercel's analysis of AI crawler traffic found that none of the major AI crawlers currently render JavaScript, a result that held across different tech stacks. Google is the notable exception: Gemini's crawler renders JavaScript because it runs on Googlebot's infrastructure.

AI agents that navigate sites on a user's behalf rely on accessibility markup too. OpenAI's own guidance states that its Atlas browser uses ARIA tags, the labels and roles screen readers rely on, to interpret page structure and interactive elements, and advises site owners to follow WAI-ARIA best practice on buttons, menus and forms.

James gave the example that made the room sit up. Review widgets like Trustpilot's are rendered in JavaScript by default. Unless you pull the reviews through the API and render them server-side, AI crawlers can't read them. His test is simple: ask ChatGPT whether it can see the reviews on your product page. It will probably tell you it can see a widget, but not what's in it. That means your best social proof is invisible to the systems now recommending products.

Chris added a warning for WordPress users. Plugins and widgets are convenient, but check that they're accessible and don't hide content behind JavaScript. You can think you're helping users while making yourself invisible to machines.

GAIN's take: It's one budget line, not two. Switch off JavaScript and look at your key pages. Whatever vanishes, AI can't see either.

3. Less traffic means every visit has to count

If AI answers mean fewer people reach your site, the ones who do matter more. That changes the job of the website. It's no longer about capturing volume. It's about earning the information people choose to give you.

James distinguished between two kinds of data. First-party data is what you observe: sessions, engagement, logins. Zero-party data is what users deliberately tell you. For a flooring retailer, that might mean asking for room size up front so the site can show a real total price rather than a price per square metre. For a small air conditioning business, his team replaced a long "fill in your details and we'll call you" form with a one-click pricing wizard. People expect the speed Amazon and Google have trained them to expect, and brands have to match it.

Chris added the condition that makes it work: respect. If someone tells you they want your email once a month, send it once a month. Use the data only for the reason it was given. Keep forms short, one question per screen if necessary, and make sure the value to the customer is obvious. Break that trust once and the data stops coming.

GAIN's take: Fewer visits make every one of them worth more. Cut the form fields, test the journey, and keep your promises about the data.

4. Governance is now a visibility issue

This was the least glamorous part of the conversation and probably the most practical. AI platforms judge your products and services by the detail on the page. If that detail is missing, you don't get recommended.

James's point was about human nature. People take the path of least resistance. If a product description or attribute field is optional, it gets skipped. So decide centrally what's mandatory: page titles, authorship, product attributes, descriptions. Then make sure your CMS enforces it, whether that's WordPress, Magento or Contentful. That decision belongs to marketing and content leaders, not to whichever developer happens to set up the field.

Chris noted that governance gets harder to maintain as teams grow and work remotely. Getting three people to follow a process is hard enough. Across an enterprise, it breaks easily. His answer is to build the checks into the publishing workflow itself, so problems are caught before content goes live rather than fixed afterwards.

GAIN's take: The quality of your product and page data is now a marketing asset, not an admin task. Set minimum content standards centrally and make your CMS enforce them.

5. Put the detail on the page, and get ready for agents

Search queries are getting longer and more conversational. People used to type "running jacket". Now they ask for a waterproof jacket for climbing Snowdon in autumn. To appear in answers to that kind of query, the evidence has to be on the page. James said brands such as Gore-Tex are moving away from minimal, photo-and-price product pages for this reason. You don't need to drown a page in copy. But what the product is for, and why it suits that use, has to be stated clearly enough for a machine to read.

There's a commercial reason too. Shopify's CFO said this month that shoppers who start their search in an LLM are 2.5 times more likely to land directly on a product page, which lifts conversion. The product page is increasingly where the customer arrives, not a step on the way there.

Then there's the bigger shift: agents that buy on the customer's behalf. James pointed to news from the week of the event, which checks out. On 21 September, Shopify announced that Meta's personal AI agent, Muse, can browse Shopify stores and complete purchases for users through Shop Pay. Not every retailer is on board. Amazon has blocked Muse, arguing retailers should decide whether third-party apps can make purchases on customers' behalf. James raised the catch for merchants. When an agent completes the sale, you keep the revenue but lose the visit, and with it the first- and zero-party data you'd otherwise collect.

Before any of this matters, you need to know whether AI systems can reach your site at all. James noted that Cloudflare had just changed its defaults, and the timing is right. From 15 September 2026, new domains joining Cloudflare have AI training and agent bots blocked by default on pages that display ads, while search crawlers remain allowed. The change also covers existing free-tier customers. Many brands will assume AI systems can see their content when a setting says otherwise.

GAIN's take: The brands that get ahead will treat their website as a source of information for machines as much as a destination for people.

The bottom line

For all the technical detail, Chris kept bringing the conversation back to one point: there's no secret formula. Good marketing is still good marketing. Accessibility markup and rendering matter, but they only help if you're relevant, specific and clear about who you serve. That's how you do well with human visitors, search engines and AI alike.

GAIN helps brands stay visible as search changes, from technical and accessibility audits to content, data and experience design. If you'd like to talk about what this means for your site, get in touch.

---