Across 3,349 answers to 25 buying questions, a clear order of AI recommendation emerges. It looks almost nothing like the market you would map from Google.

An industry study by Lil Big Things.

Want this study as a designed PDF?

Enter your email for the full PDF and the raw dataset. The study stays free to read either way.

You will also get each new study. Unsubscribe anytime.

Healthcare generates more data than any other industry — and strands most of it in incompatible EHRs, legacy databases, and dead formats. The vendors solving this sell overlapping capabilities under different banners: EHR data migration, application retirement and archival, interoperability, real-time streaming, and research enablement. Buyers rarely name the need the same way twice.

That linguistic sprawl is the defining trait of this category, and it is why AI assistants behave so differently here than a keyword search does. Ask a search engine for "EHR data archiving" and you get archival specialists. Ask an assistant "how do we move data between our EHRs?" and you get the biggest, most-written-about names in health IT — whether or not those names actually sell the service.

Two competitive sets exist. The search set — who ranks on Google for the category keywords — is specialists. The answer set — who AI names when buyers describe the problem in their own words — is incumbents. This report measures the answer set, because that is the room where the category is now being decided.

Every prompt was split by buying stage. The category behaves very differently at the top of the funnel than at the bottom — and the reshuffle is where deals are won.

At the awareness stage — "what platforms handle healthcare data?", "what tools move data between EHRs?" — the field is wide and generous. Assistants list many names, and even tail brands get surfaced. At the evaluation stage — "how do I choose?", "what does it cost?", "which is best for security?" — the field collapses onto the names with the deepest comparative paper trail.

Across the category, the incumbents carry their presence intact from awareness into evaluation. The specialists mostly do not. Hart is the clearest example: present in 26 awareness-stage prompt variants, it survives in just 2 evaluation-stage ones — a near-total dropout at the exact moment a buyer asks the machine to help them decide.

Awareness visibility does not carry to evaluation on its own. Evaluation answers are built from comparative, third-party evidence — best-of lists, buyer guides, analyst notes, pricing discussions. Brands without that corroboration get listed early and dropped late. In this category, that is the difference between the incumbents and everyone else.

There is no single "AI answer." Presence varies dramatically by assistant, and the variation is not noise — it tracks how each engine builds its answer.

Sort the engines by how they answer and the split is stark. Engines that search the live web in real time — Copilot, Google AI Overviews, Gemini — surface specialists far more readily. Engines that answer primarily from trained memory — ChatGPT, Perplexity, and Claude — default hard to the incumbents.

Hart illustrates the pattern clearly. Its per-engine share: Copilot 9.2%, Google AI Overviews 7.3%, Gemini 5.5%, ChatGPT 2.8%, Perplexity 1.9%, Claude 0.0%. One specialist scored zero across 425 Claude answers while reaching 9.2% on Copilot. The same brand, on the same category, invisible in one engine and meaningfully present in another.

A brand is not fit for the job barely moves the memory engines — they name who they were trained to know. But the retrieval engines can be influenced in weeks with the right content and citations. The category is split into a fast lane (retrieval) and a slow lane (memory), and a brand strategy depends entirely on which lane it is fighting in.

Hart per-engine presence rates, illustrating the retrieval vs. memory engine split. Same brand, same category, completely different result depending on the engine.

Every appearance maps to a real question. Read across the six question themes and a clear pattern emerges: who shows up narrows sharply as questions move from open exploration toward comparison and decision.

Epic has consistent presence (solid) across all six themes. Rhapsody Health and InterSystems are consistently present across integration, interoperability, and storage themes but fall off in research and analytics. Arcadia is analytics-strong. Hart shows partial presence in storage and real-time prompts but is rare or absent in security, compliance, and evaluation queries.

The pattern holds across the category: the broader and more exploratory the question, the more open the field; the more it demands a comparison or a decision, the harder it collapses onto Epic and the integration tier. The winnable ground for any challenger is the cluster of specific, capability-led questions — "historical + real-time," "data integrity in transfer," "consolidate from many sources" — where a precise, well-documented answer can outrank a famous but generic one.

Specificity beats fame on narrow queries. The brands that win contested prompts are the ones whose documentation is unusually exact about a capability — and whose exactness is echoed on third-party pages. Generic breadth loses these; precise depth wins them.

When an engine searches to answer a category question, the cited URLs reveal the machinery beneath the entire leaderboard. Approximately 94% of cited URLs are third-party — aggregators, journals, vendor round-ups, directories. About 5% point to competitor or vendor-owned domains. The specialist brand own domains account for roughly 1% of the evidence base for their own category.

On Perplexity specifically, third-party citation share exceeded 97%. A Perplexity answer to an open "historical + real-time" query on 11 Jul 2026 cited 10 sources, all third-party, and named no vendor by name.

AI assistants do not answer category questions from a vendor website. They answer from the consensus of the open web, then — if a brand is famous enough — from memory. This is why incumbents win: two decades of independent writing about them is the evidence base. A brand own content, however good, is weighted far below independent corroboration.

Reading the raw responses side by side exposes the real structure of AI visibility in this category. An assistant names a brand through memory or through retrieval — and the two doors need entirely different keys.

Door 1 is memory. Opened by fame, over years. Engines answering from training data name brands they "know." They reach for the names with a massive footprint of independent writing — Epic, InterSystems, Rhapsody, the hyperscalers. Challengers are not in that memory yet, and cannot be for quarters. A Claude response to an open category question reads: "Epic and Oracle Health... Rhapsody, Mirth, InterSystems HealthShare... AWS HealthLake, Google Cloud Healthcare." Zero citations. Pure recall. No specialist named.

Door 2 is retrieval. Opened by content, in weeks. Engines that search live name brands they find in top sources. A well-documented specialist can surface here — but only if it appears in the third-party pages the engine retrieves, not just on its own site. A Perplexity response to the same category pulled 10 third-party citations, zero vendor-owned. Retrieval happened; the specialists simply were not in the retrieved set.

Incumbents win both doors: they are famous (memory) and heavily cited by third parties (retrieval). Challengers win neither by default. But the retrieval door is openable now, through third-party placement and precise content — while the memory door opens slowly as that same presence trains into the models. The whole challenger playbook follows from this asymmetry.

The ten brands sort into four positions in the AI answer. Each faces a different door problem — and a different path forward.

The Default (Epic, 34%, both doors open): The name AI reaches for first, in memory and retrieval alike. Owns the evaluation stage because the comparative web is saturated with it. Effectively uncontestable head-on — the strategic error would be attacking it directly.

The Integration Tier (Rhapsody Health, InterSystems, Arcadia — 10 to 11%, memory-strong): Well-known integration and data platforms with deep independent documentation. They carry from awareness into evaluation intact. This is the realistic ceiling a challenger competes toward — not Epic.

The Mid-Field (Innovaccer, Infor/Cloverleaf, Health Gorilla — 4 to 6%, uneven): Recognized names with partial coverage — strong on some topics, absent on others. Their presence rises and falls with how much third-party writing exists on each specific question.

The Specialists (Hart, Datavant, Imprivata — 1 to 3%, retrieval-only): Purpose-built for the actual job, described accurately whenever named, but rarely summoned. They surface on retrieval engines and vanish on memory engines. The category clearest opportunity — and its clearest cautionary tale about fitness not equalling visibility.

Hart is the sharpest illustration of the specialist archetype. It is built precisely for healthcare data accessibility — migration, archival, streaming, research enablement — and when AI names it, the description is accurate and the sentiment runs effectively 100% positive with zero negatives. Yet it sits at 3.3% overall, 0.0% on Claude, and near-absent at the evaluation stage. It is the best-described, least-summoned brand in its category — invisible in memory, thin in retrieval, flawless in sentiment. Its problem is not reputation; it is frequency. And frequency is the more fixable of the two.

This study ran 25 non-branded seed questions across 6 AI engines between 16 Jun – 13 Jul 2026, capturing 3,349 machine-generated answers. Category structure and the organic competitive set were derived from Ahrefs. AI-visibility metrics were measured in Scrunch AI.

Prompts are non-branded, written the way a real buyer describes the problem in their own words. We distinguish between awareness prompts and evaluation prompts. Figures are directional estimates from third-party measurement, not audited results.