A marketing director at a $150M medtech company is asking Perplexity to compare healthcare marketing agencies right now.
She hasn’t visited your website. She hasn’t clicked anything. She doesn’t appear anywhere in your CRM. She’s just sitting somewhere with a coffee, asking an AI tool to do the early-stage research she used to do herself across six browser tabs and three industry reports.
By the time she fills in a contact form, the shortlist is already formed.
And if you’re not in it, you were never in the running.
This isn’t a prediction. This is already happening, and most marketing teams are measuring the wrong things to see it.
The research moved and nobody told your analytics.
Forrester’s 2026 Buyers’ Journey Survey of nearly 18,000 global business buyers found that generative AI and conversational search now rank as the most meaningful source of purchase information for enterprise buyers. More meaningful than vendor websites. More meaningful than direct contact with sales teams.
94% of B2B buyers now use AI in their purchase process. The proportion who name it as their primary research source more than doubled in twelve months.
What this means in practice is uncomfortable.
By the time an enterprise buyer appears in your marketing funnel, they’ve already formed strong opinions about which vendors are credible, which are relevant, and which are worth their time. Those opinions were formed inside systems you can’t see, can’t retarget, and can’t influence through most of the things your demand generation programme is currently doing.
The marketing funnel still exists. It just has a new phase at the top that precedes all of it.
Call it the AI research phase. Call it the dark funnel. The name matters less than the implication: your organisation needs to exist credibly inside AI-generated answers, or it doesn’t exist at all for a growing proportion of the buyers you most want to find you.
Regulated industries have a specific version of this problem.
The AI research phase affects every B2B organisation. For healthcare, financial services, enterprise technology and education brands, it creates an additional layer.
AI platforms apply their own quality filters before citing content about regulated industries.
For healthcare specifically, Google’s YMYL classification (Your Money Your Life) means health content is evaluated to a higher standard of expertise, authoritativeness and trustworthiness before it’s used as a citation source. The same logic applies to financial services content under FCA fair presentation standards, and to pharmaceutical content under ABPI and FDA guidance.
Here’s what that means in practice.
A healthcare marketing agency whose content doesn’t demonstrate clinical credibility, named authorship with verifiable credentials, and accurate sourcing is less likely to appear in AI-generated answers about healthcare marketing regardless of how much content it’s published. An enterprise buyer who asks an AI tool ‘which marketing agencies specialise in MedTech’ will receive answers drawn from sources that AI engines have assessed as authoritative on those topics. If your content doesn’t meet that bar, you’re invisible regardless of your actual capability.
There’s also the accuracy problem, which is more specific to regulated industries.
An AI tool that misrepresents a healthcare agency’s compliance capabilities, or cites outdated information about a financial services firm’s regulatory expertise, creates reputational risk for the firm being described. Getting cited accurately is as important as getting cited at all. These are not the same thing.
What enterprise buyers are actually doing in AI tools.
Before getting to the mechanics, it’s worth being precise about the kind of prompts enterprise buyers in regulated industries actually run, because they’re quite different from generic search queries.
They’re not searching for your name.
They’re searching for your category.
‘Which marketing agencies have deep experience in regulated healthcare markets.’
‘Compare digital marketing firms specialising in MedTech and Pharma.’
‘What should I look for when choosing a marketing strategy consultant for a medical device company entering the US market.’
Category and evaluation queries. Being cited in them requires that AI engines understand your organisation as a credible authority in that specific category, not just that they know your name.
They’re also using AI to validate signals they’ve already received. A buyer who encountered your firm at a conference or through a referral will run a prompt to check before acting on it. ‘Is [agency name] credible for healthcare marketing?’ The AI response to this validation query is now a trust checkpoint in the sales process. If the response is thin, outdated or contradicts your positioning, the buyer’s confidence in the initial signal decreases.
And here’s the one most organisations have genuinely not clocked yet.
68% of enterprise buyers use Microsoft Copilot. More than a third use a private instance behind their company’s firewall. That tool draws on Bing’s web index alongside internal company documents. An enterprise procurement manager at a hospital group can ask Copilot to compile a vendor briefing document without ever leaving Microsoft Teams. The briefing is assembled from web sources Bing has indexed as authoritative.
Your organisation either appears in that briefing or it doesn’t.
There is no retargeting pixel that can intervene.
What determines whether AI engines cite you.
The signals that determine whether your content is cited in an AI response are specific and addressable. This is good news. It means there’s something to actually do here.
Structured data and schema markup tell AI engines what your organisation is, what it does, who it serves and what credentials it holds. FAQPage schema that directly answers the questions your buyers ask AI tools. Service schema that describes your specialisms accurately. Organisation schema with complete, consistent data. Without these, AI engines have to infer your category, capabilities and credentials from unstructured text — and they frequently infer incorrectly.
Named authorship with verifiable credentials is the single most important signal for regulated industry content. AI models assess health and financial content more favourably when it’s attributed to named individuals with verifiable expertise. A thought leadership article on MedTech marketing strategy attributed to the team is a weaker citation candidate than the same article attributed to a named person with a demonstrable track record in healthcare. Research published in early 2026 found that AI models systematically under-cite content without personal attribution. This is not a small difference.
Topical authority, the depth and breadth of content coverage on a specific subject, is how AI engines assess whether an organisation is genuinely expert in a category. An agency that has published comprehensive, accurate content across medical device marketing strategy, healthcare SEO, YMYL compliance, MedTech PR and healthcare content marketing has demonstrated the topical depth AI engines recognise. An agency with one generic healthcare page has not.
Earned media in authoritative publications is the signal most AI visibility strategies miss. AI engines don’t only assess the content on your own website. They synthesise information from across the web, and the sources they weight most heavily are third-party publications – trade press, industry journals, recognised media. Brands with consistent coverage in sector-authoritative publications show AI citation rates four to seven times higher than brands with equivalent owned content but no third-party presence.
The PR programme and the AI visibility programme are, in 2026, the same programme.
Organisations that treat them as separate are building half a distribution engine.
Content freshness. AI-cited content is 25.7% fresher on average than organic search-cited content. Regularly updated existing content and not just new content performs better than content that hasn’t been touched in twelve months. For regulated industries, outdated regulatory information is a citation risk as much as a ranking one.
The compounding distribution engine.
I want to be direct about something, because a lot of what’s being sold in the AI visibility space right now is a technical fix.
Schema implementation. Prompt engineering. A new GEO audit tool. All real, all useful, all insufficient on their own.
The brands that consistently appear in AI-generated answers about their category are not there because they found a trick. They are there because they’ve built genuine topical authority, earned media presence and a structured content ecosystem over time. There is no shortcut for that.
The compounding engine has four parts.
Owned content built for AI citability. Comprehensive, accurate, deeply specific content that addresses the actual questions enterprise buyers ask AI tools when researching your category. Not blog posts optimised for generic keywords. Not thought leadership that reads as marketing. Structured, authored, evidence-based content that directly answers evaluation questions: what is the best healthcare marketing strategy for a medtech company entering the US market, what should a CMO look for when choosing a financial services marketing agency. This.
Earned media in authoritative sector publications. Every placement in a recognised healthcare, technology, finance or education trade publication is a citation signal that AI engines weight significantly above owned content. A bylined article in Digital Health. A quoted opinion in Healthcare Manager. A case study referenced in MedCity News. These are not just PR wins. They are AI citation assets.
Structured entity definition across the web. AI engines build their understanding of your organisation from multiple sources simultaneously: your website, third-party publications, directory listings, LinkedIn, Google Business Profile. Consistency across all of these allows AI engines to cite you confidently and accurately. Inconsistency produces inaccurate or absent citation. It’s a simple problem with a boring solution.
Prompt testing as measurement. Unlike traditional SEO where GSC gives you impression and ranking data, AI visibility requires direct measurement. Open ChatGPT, Perplexity, Google AI Overviews and Microsoft Copilot. Run the prompts your buyers run.
‘Best healthcare marketing agencies UK.’
‘MedTech marketing strategy consultants.’
‘Marketing agency for regulated industries London.’
Document whether you appear, how you’re described, whether the description is accurate, and which competitors appear when you don’t. This isn’t a one-time exercise. It’s a monthly discipline.
The advantage nobody is talking about.
Here’s the thing about regulated industries that I find genuinely interesting.
The compliance and accuracy requirements that make marketing in healthcare and financial services harder than other sectors, the clinical review, the named authorship, the regulatory sign-off, the accurate sourcing – are exactly the same requirements that AI engines apply when assessing content quality.
If you’ve been doing this properly, you’ve already built the content foundations that AI citation requires.
You’re not starting from scratch. You’re recognising that what you’ve been doing for compliance reasons also produces the signals AI visibility needs.
The gap is usually in three specific places. Structured data implementation including FAQPage schema, Service schema, and Organisation schema is often missing even when the content is strong. The third-party citation programme is often underdeveloped relative to the quality of the owned content. And the measurement practice, regular prompt testing, is almost universally absent because it’s a new discipline that didn’t exist eighteen months ago.
Those three gaps are real, but they’re all closeable.
Where to start?
Run a prompt audit. Think of it as the entry point to any generative engine optimization programme worth building.
Open Perplexity, ChatGPT and Google AI Overview. Run ten prompts that mirror how your buyers research your category. Document the results honestly.
When a buyer runs a category research prompt in your sector, do you appear? When you’re cited, is the description accurate? What’s the gap between your AI citation presence and your nearest competitors? What are the highest-priority citation opportunities in your specific sector?
The buyers who will be your most valuable clients in the next twelve months are already running these prompts.
Whether you appear in the answers they receive is determined by decisions you make about content, earned media and structured data now. Not by the quality of your response when they eventually find their way to your contact page.
And that’s the part that’s uncomfortable to sit with.
Because by then, the decision is probably already made.