Logo

The short version

Websites still matter, but the job they do is changing.
Search traffic from Google is getting harder to earn as AI tools answer more questions directly, but the visitors who do arrive from AI referrals are more engaged and more likely to convert.
Winning now means being a source that AI systems trust and cite, which comes down to clear structure, evidence and expertise rather than keyword rankings alone.
Below, I go through what we covered in our recent webinar, including a live example from RCOT, excellent input from Manifesto and what it means for your organisation.

At the start of our recent webinar, Will your website still matter in an AI world, I asked the audience a simple poll:

how prepared do you feel for the shift to AI driven search?

The results were telling: almost nobody felt well prepared.

The majority said they were aware of the need to take action but hadn't yet done so. And, as Dominika Cechova, Digital Marketing Strategist at Manifesto, pointed out, these are the same questions the sector has been asking for two years. So, what's stopping people from moving?

Alongside Anna Faithfull, Head of Digital Channels at the Royal College of Occupational Therapists (RCOT), and Dominika from Manifesto, we set out to explore this question.

If you missed it, you can watch the recording here. If you were there, what follows picks up where we left off, including some reading I'd recommend alongside it.

How AI search is changing traffic for nonprofits

Let me start with the data, because I think it puts the urgency into perspective.

Search changed more in the last two years than in the previous 20.

Today, 68% of digital journeys start with a search query, but only 40% of Google searches result in a click through to a website. That figure comes from Sparktoro, Aura and SEMrush's 2024 report, and I'd expect it's moved further since then.

According to M+R's research across 17 US nonprofits, these organisations saw double digit growth in organic search traffic in the months before Google widened its rollout of AI Overviews in March 2025. By August, that had reversed into a 13% year on year drop in organic search visitors, even though searches for those same organisations' names had risen by almost 19% over the same period.

More people looking, but fewer people landing.

AI search tools surface a small number of trusted sources and synthesise answers directly. If your content isn't structured, clearly attributed or recognisably authoritative, you simply don't appear.

Why AI referral traffic is worth more than it looks

When someone does arrive at your site via an AI referral, they are a significantly better visitor.

Based on SimilarWeb data, referrals from AI platforms grew 778% in a single year. Visitors arriving from ChatGPT spend an average of 15 minutes on site, roughly twice the dwell time of a typical Google referred visitor. They view more pages per session (12 versus Google's 9), and they convert at higher rates. The reason is straightforward: they've already had a substantive exchange with an AI tool, so they've formed a clearer picture of what they need and arrive with intent.

So the shift isn't simply about losing traffic; it's about the nature of the traffic changing.

Fewer, better visits from people who are genuinely ready to engage.

A real example: how RCOT adapted to AI search

Rather than deal in the abstract, we spent the core of the session on a real example.

Anna Faithfull joined us to talk through the work RCOT has been doing with their digital agency, Manifesto, following a full website replatforming last year.

When Google AI Overviews arrived in 2025, positions one to three in search results, which everyone had been working hard to achieve, were suddenly pushed below the fold overnight.

Anna described it : "You could be working really hard on your search engine rankings, and they just became less relevant."

RCOT's response was methodical.

They started by looking at external search and discoverability. They audited their content, identified inconsistencies in how questions and answers were being presented across the site and released a new FAQ component: a structured, properly marked up accordion with question and answer fields, built as a global library to consolidate content and make it genuinely parseable by AI systems.

Alongside that, they worked with Manifesto to introduce a hybrid AI-powered search function on their own site, running alongside traditional search in a live A/B test.

The same search term ("what is occupational therapy?") returns meaningfully different results: one experience surfaces a structured AI-generated summary alongside tabbed, theme-grouped results that front-load the most relevant content; the other runs the original keyword-based search.

Early indicators are positive, though Anna was refreshingly candid that they haven't yet reached statistical significance and that some findings are already flagging areas for further testing.

The metrics they chose to track are worth noting too. Time on page was rejected as too ambiguous. Instead: average result position clicked and searches per session. Simple, purposeful and directly tied to whether the search function is actually serving user intent.

What struck me about RCOT's approach is that it embodies something Dominika said:

“ The worst thing you can do is scrap what's working and jump on the latest tactics.”

About 75% of the underlying strategy is the same, so it's an evolution rather than a replacement.

What helps your website get found by AI systems

The goal has shifted from rankings to recognition.

You are no longer trying to rank number one; you're trying to be the source that AI systems cite and trust. That means being clearly and consistently identified as an authoritative voice on your cause or field.

Your content has one audience, but speaks two languages. One for us humans and the other for AI.

Write for humans. That is still the primary job.

But AI systems respond well to structure, credentials, statistics, citable references and schema markup. These things support great content; they don't substitute for it. As Dominika put it, if your digital presence is your shopfront, and nobody can find it, what's the point?

Original, useful content outperforms volume.

AI-generated content carries real risk. AI can detect AI-written content, and thin content will harm your rankings. What gets cited is specific: your research, your case studies, your expertise, your voice.

Your digital presence is bigger than your website. 95% of AI citations come from sources other than the cited organisation's own site. Your credibility is shaped by what directories, publications, review platforms and external sites say about you.

A traffic drop does not automatically mean you are failing.

Fewer visits may mean that AI is answering the question before people click. Look at depth of journey, dwell time, conversion, not just traffic volume. There are tools that can now show you where and how often you're being referenced in AI platforms; it's worth starting to track this.

And one more thing: humans aren't going anywhere.

AI may be doing more of the filtering, but people come to your website to connect, to be moved, to decide. Tone of voice, values, narrative and emotional resonance all still matter enormously. You can win on technical optimisation and still lose on the thing that makes someone join, donate or get in touch.

Will websites still matter?

We ended with the big question an attendee put into the chat:

If websites become less central, what's the new digital front door?

I don't think anyone can answer that with certainty.

What we can say is that for organisations in the charity, membership and NFP space (where content is substantive, nuanced and built on genuine expertise), websites will continue to matter, because humans will continue to want more than a summary can give them.

As Dominika put it: we still use paper even though we have computers.

The path to your website is changing.

How it's presented to AI systems will matter more over time. But the organisations that will do best are the ones that are genuinely authoritative, consistently credible and structurally well set up, because those are the same things that make a website useful to a human.

What to read next

If this question has prompted you to think harder about your digital strategy, here are three pieces of reading I'd suggest alongside it. They sit directly with the themes we covered, and I've included a word on why each one is worth your time.

  • Optimising for Generative AI Features on Google Search:
    Google's own guidance, and it's more accessible than you'd expect. If you're trying to make the case internally for why your content strategy needs to evolve, going to the source is a stronger starting point than citing someone else's interpretation of it.
  • SEO's New Goal in 2026: Recognition, Not Rankings:
    This captures something central to our session: visibility in an AI world is increasingly about authority, trust, and how well your organisation is understood across the web — not just where you appear in a list of results. Useful reading for anyone thinking about digital strategy right now.
  • It Works Until It Doesn't: AI Content Strategies That Backfire:
    A useful counterpoint to the more optimistic AI content narratives. Scaling content with AI can look like a smart move, until it isn't. Worth reading before doubling down on any AI-first content approach.

Taken together, they reinforce what came through in our session:

The organisations that will do best in AI-driven search are the ones that are genuinely authoritative, consistently credible, and structurally well set up.

That's harder to shortcut than it sounds, and it requires thinking about your technology, data, and content as connected problems rather than separate ones.

If you'd like to talk through what any of this means for your organisation specifically, I'm glad to make time for a conversation.