Our clients publish online because they want their research to be read, shared and acted on.
They want policymakers to rethink a position, businesses to change their behaviour, or the public to see an issue in a new light. Impact is the goal.
But impact depends on being found.
And the way people find information online is changing fast. Traditional search is no longer the only gateway. Increasingly, audiences turn to AI assistants like ChatGPT for instant answers.
That shift raises urgent questions: Will our work still be found online? Will it still carry weight? Will the right people still see and act on it?
To understand what this means in practice, we analysed the analytics dashboards of three major research organisations. Each is authoritative in its field, each publishes high-quality primary research, and each seeks broad reach.
Let’s call these organisations Alpha, Bravo and Charlie. The names are fictional, but the data is real.
Looking at AI referrals across a similar timeframe:
- Alpha: ~5% of all visits arrived from ChatGPT referrals.
- Bravo: ~1% arrived from ChatGPT.
- Charlie: negligible traffic from ChatGPT (and similarly low from other assistants like Perplexity or Claude).
Why do we see such different outcomes for sites with a similar mission, authority and calibre of research?
The variance isn’t about brand power or research quality.
Our research suggests it's more about how the research is presented on the web.
Alpha: The AI sage
Alpha lays out its findings in a way that is easy for both humans and machines to digest. Each research page includes:
- A short abstract in plain HTML.
- Headings for methodology, findings, and conclusions.
- Key statistics are stated on the page, not buried in downloads.
This makes Alpha highly “quotable” for a large language model (LLM), such as ChatGPT. The assistant can easily lift a headline fact, attribute it to Alpha, and direct users to the right page for more.
The net result is a strong share of traffic from ChatGPT. In AI’s brave new world, Alpha is seen as a source of ready-to-use, verifiable knowledge.
Bravo: The human storyteller
Bravo also produces outstanding research, but its website favours narrative storytelling. Articles highlight the importance of the findings, profile the researchers, and explain the relevance to society.
This approach is excellent for engaging human readers, but less useful for an AI looking for a specific claim or statistic.
Bravo’s magazine-style of thoughtful features with a strong editorial voice offers a polished experience for humans. For machines, it creates friction.
While only gaining a fraction of Alpha’s ChatGPT referrals, it still dwarfs social traffic to the site. Bravo’s authority isn’t in question. It’s just that robots find its outputs harder to cite.
Charlie: The PDF vault
Charlie has perhaps the richest store of data of all. But almost all of it is locked away inside PDF reports.
The accompanying webpages are usually short with a summary paragraph introducing the dreaded “DOWNLOAD” button.
For both robots and people, this is hard work. ChatGPT prefers HTML text it can parse, anchor, and cite directly. This means models often overlook it in favour of more “machine-friendly” sources.
The result is negligible traffic from ChatGPT, despite world-class research. The robot assistant will default to sources that are easier to read and cite.
Why does this matter?
Visitors arriving from ChatGPT are not fleeting or casual. They come with purpose. They're looking for the depth and expertise that each of these clients can offer.
Our research suggests that, on average, they view more pages than those arriving from social media and sometimes, even more than those coming via organic search.
They also stay for longer. Time on site from AI referrals is consistently healthy. A sign that these users are not just grazing the surface but digging deeper into your content. They come for an answer, find it, and then want to learn more.
There’s a common misconception that traffic from AI assistants is somehow less worthy and that people are outsourcing the hard work of reading to the machine.
But the data we analysed shows the opposite. When AI sends visitors through, they are highly qualified: they know what they’re looking for, they trust the source, and they engage deeply with what they find.
In other words, these aren’t low-value visits. It’s highly engaged, motivated traffic that matches exactly what research organisations want.
These new audiences are discovering your work through new channels, then taking the time to read, reflect and act on it.
How to check your traffic from ChatGPT
It’s important to know how many visits you have from AI assistants. The simplest way to do this is through Google Analytics (GA4).
Here’s a quick step-by-step guide:
1) Open your reports
- In GA4, go to Reports > Acquisition > Traffic acquisition.
- Switch the primary dimension to Session source/medium.
2) Look for ChatGPT referrals
- Scan the table for sources such as chat.openai.com / referral.
- You may also see traffic from other assistants like perplexity.ai or claude.ai.
- Export or segment this traffic so you can evaluate it separately.
3) Assess relative value
- Compare pages per session and average engagement time for ChatGPT referrals against your other channels (organic search, social, direct).
- Look at conversion events (downloads, sign-ups, contact clicks). Do AI visitors complete them at a higher or lower rate?
4) Interpret what you see
- If AI visitors are a healthy percentage of your traffic and highly engaged, you may be more like Alpha, the citable researcher.
- If you’re getting some AI traffic but less than social or other channels, you may be closer to Bravo. As a storyteller, your content resonates but lacks the citable scaffolding machines prefer.
- If you see negligible AI referrals despite strong research outputs, you may be Charlie, a PDF vault with content that is valuable but not easily surfaced by AI because it is locked away in downloads.
The exercise is simple but revealing. With just a few clicks in GA4, you can benchmark your organisation against this new reality and start to see where you sit in the Alpha–Bravo–Charlie spectrum.
10-point plan to make the most of your AI traffic
- Measure it: Do you know how much of your current traffic comes from ChatGPT and other assistants?
- Value it: Are AI visitors spending longer on site and reading more pages than those from social or direct traffic?
- Summarise it: Are your key findings easy for humans and machines to grasp at a glance?
- Show your workings: Do you surface methods, limitations and sources in ways that build trust and credibility?
- Unlock the vault: Is vital content trapped in PDFs when it could be discoverable in HTML?
- Make it citable: Can each claim be quoted, anchored and linked directly to your page?
- Share the evidence: Do you provide data, charts or structured outputs that others can reuse and reference?
- Be machine-friendly: Is your site fast, clean and readable for lightweight browsers as well as people?
- Guide the visitor: Do you help AI-referred users take the next steps, such as subscribing?
- Prune and polish: Are you maintaining a clear, authoritative library of research, free of thin or duplicate pages?
Getting this right isn’t about ripping up your website and SEO strategy. It’s about understanding where you stand today, and where you’d like to be tomorrow.
Are you Alpha, Bravo or Charlie?
If you’re curious about how your organisation can make the most of this new wave of AI-driven discovery, we’d be glad to share what we’ve learned along the way.
