Your digital experience has three readers now

For twenty years, digital teams optimised for two readers. The person, who needed to understand and trust you. And Google, which needed to crawl and rank you. Entire disciplines grew around each: UX and content design for the first, SEO for the second. Organisations got reasonably good at both.

There's a third reader now. When someone asks ChatGPT, Claude, Gemini or Google's AI Overviews which bank suits their situation, which broker to call, which product to shortlist, an AI assistant reads the available evidence and answers on their behalf. The person may never see your website at all. They see what the assistant says about you, or more often, doesn't.

The failure is specific, not general

Here's what makes this urgent rather than theoretical. When I tested a set of Australian financial services brands against the major assistants, factual queries mostly came back fine. Ask about a specific named organisation and the assistant retrieves accurate details. The brands looked at that and relaxed.

Category queries were a different story. Ask the question an actual customer asks, best option for my occupation, my industry, my situation, and the assistants confidently named competitors. Not because the competitors were better. Because the competitors had published the structured, crawlable, plainly worded evidence that lets an assistant justify a recommendation, and the invisible brands hadn't. An assistant can't cite what it can't parse, and it won't recommend what it can't cite.

That's the pattern to internalise: factually accurate, categorically invisible. Your brand can pass every fact-check and still never be the answer.

Why your existing playbook doesn't cover this

The instinct is to treat this as SEO with a new name. It mostly isn't. Google rewards many things assistants ignore, and tolerates many things assistants can't handle at all. Google executes JavaScript; most AI crawlers don't, so an experience rendered client-side can rank beautifully and serve an assistant an empty page. Google rewards domain authority accumulated over decades; an assistant assembling an answer tonight rewards clear claims, comparable structures, and pages that map to the actual question asked.

The unit of competition has changed too. In search, you competed for a ranking on a results page the person still had to read. In assisted discovery, you compete to be the answer itself. There is no page two. There's barely a page one.

What to actually do

The good news is that the fix is unusually cheap relative to most digital investments, because it's mostly about structure and candour rather than technology. Three moves, in order.

First, fix the citation layer. Make sure what's true about you is stated somewhere plainly, in server-rendered HTML, in words a machine can lift without interpretation. Eligibility, coverage, pricing logic, who you serve. Most organisations have this information; they've just buried it in PDFs, tabs and marketing prose.

Second, build for the category questions. People don't ask assistants for your brand; they ask for their situation. Hub-and-spoke content that answers situation-level questions, honestly and specifically, is what earns the citation. This is exactly the structure I build commercially at Rankline, and it works because it gives the assistant something defensible to point at.

Third, only then touch the journey. Being recommended creates a new kind of arrival: a visitor who lands mid-task, pre-sold by a machine, expecting the experience to confirm what they were told. If your join or enquiry flow contradicts the assistant's summary, you convert the click into distrust.

The uncomfortable part

The third reader punishes vagueness. Assistants recommend what they can verify, which means marketing language that says nothing checkable is now a commercial liability, not just a style problem. The organisations that win the next few years of discovery will be the ones willing to state plainly who they're for, who they're not for, and why. That was always good practice. Now it's infrastructure.

Nathan Haslewood is a digital experience leader in Melbourne. This site practises what the essay preaches: view any page as each of its three readers from the homepage, or see how it's built.