While AI founders chase crowded categories such as AI SDRs, AI customer support, and AI coding, one less obvious but commercially valuable direction is being quietly occupied: AI assistants for academic research.
Elicit, an AI research assistant incubated from the nonprofit Ought in 2022, is spreading through academic and research workflows at notable speed. Its homepage highlights the scale directly:
More than 5 million researchers, more than 125 million indexed papers, and R&D teams at global pharmaceutical companies.
This is not a story about AI being powerful in the abstract. It is a story about how to productize AI correctly inside a vertical workflow.
01 It Did Not Build “Smarter AI”; It Compressed the Most Painful Workflow
Before building the product, the Elicit team spent serious time understanding one question: what is the most painful part of a researcher’s work?
The answer was not “not enough information.” The answer was “too much information, and screening it is too slow.”
A typical literature review looks like this:
- Search research databases such as PubMed or Google Scholar.
- Browse hundreds or thousands of results and manually screen relevant papers.
- Read abstracts to judge relevance.
- Read full papers to extract key data.
- Organize the findings into a table or review.
For a graduate student, this process can easily take two to four weeks.
Elicit did something simple and smart: it compressed that whole process into an end-to-end product.
The user enters a research question. Elicit searches across 125 million papers, screens the most relevant results, extracts key data into a table, and generates a summary report.
This is not a small feature wrapped around an LLM. It is a full product redesigned around a specific workflow.
Productization score: five stars
The key distinction is that Elicit was not built to demonstrate AI capability. It was built to replace a real workflow.
02 Compliance Is the Real Moat
In May 2026, Elicit made a decisive move: it added support for the PRISMA 2020 systematic review standard.
If you are outside academia, PRISMA may not mean much. But understanding this move explains Elicit’s commercial logic.
PRISMA, the Preferred Reporting Items for Systematic Reviews and Meta-Analyses, is a gold standard in medicine, public health, and related fields. A systematic review that does not meet PRISMA expectations may fail peer review.
Elicit not only supports PRISMA. It also published evaluation results:
- Search recall: 95%
- Abstract screening accuracy: 97%
- Full-text screening accuracy: 99%
- Data extraction accuracy: 96%
Those numbers came from tests against 994 Cochrane systematic reviews.
The smart part of this strategy is that Elicit is not trying to acquire customers by saying “AI is better than humans.” It is building trust by saying “AI can meet the standard your field already trusts.”
In research, authors are cautious about directly citing a ChatGPT-generated review because hallucination risk is real. But an Elicit-assisted PRISMA-compliant systematic review can be treated as a work product inside a serious research process.
Commercial moat: five stars
This is the fundamental reason Elicit can sell into pharmaceutical R&D. The product is valuable not merely because it uses advanced models, but because its output can fit into real academic and regulated work.
03 Commercialization in Three Steps: From Tool to Platform
Elicit’s commercialization path can be understood in three stages.
Stage one: freemium acquisition, 2022-2024
The core search experience was free, giving researchers a low-friction way to try the product and build a habit. Academic markets are slow-moving, but loyalty can be high. Once a lab gets used to a tool, migration cost becomes meaningful.
The result: 5 million users, largely driven by organic growth.
Stage two: paid subscription, 2024-2025
Elicit introduced Plus, Pro, and Enterprise tiers. Research teams that need data extraction, advanced reports, and collaboration hit the paywall. This is classic SaaS packaging: the basic workflow is free, but the higher-value work is paid.
Stage three: platformization, 2026 onward
This is the most important stage to watch.
In March 2026, Elicit launched an API. Third-party developers can now embed Elicit’s research capabilities into their own products.
In December 2025, it introduced Research Agent workflows, letting AI execute multi-step research tasks.
In July 2025, it connected to more than 545,000 ClinicalTrials.gov records.
In April 2026, the company wrote that it had held more than 500 deep conversations with top pharmaceutical companies. The direction is clear: enter drug R&D workflows.
Why is the API important?
Because an API moves Elicit from a personal tool into infrastructure. A pharmaceutical company can integrate Elicit into internal systems. Once that happens, it is no longer something an individual user casually switches away from. It becomes part of enterprise workflow stickiness.
04 What Makes It Different from General AI?
A natural question is: ChatGPT and Claude can also search, summarize, and reason over papers. Why would researchers use Elicit?
The answer is workflow depth.
- ChatGPT can summarize a paper.
- Elicit can help complete a systematic review while aligning with PRISMA.
- ChatGPT answers questions.
- Elicit replaces a workflow.
The analogy is simple:
- General AI tools are like a Swiss army knife. They can do a little of many things.
- Elicit is like a specialized surgical instrument. It does one thing, but with professional depth.
For anyone who thinks AI entrepreneurship is just “wrapping ChatGPT,” Elicit is a strong counterexample. Valuable AI products do not merely package model capability. They redesign a workflow.
05 Three Lessons Builders Can Copy
Lesson One: Understand the Workflow Before Adding AI
Elicit did not add an “AI search” label to an API. It studied the full research workflow, found the highest-friction pain point, and rebuilt the product around screening and extraction.
Concrete practice: interview target users deeply. Elicit has described more than 500 conversations with pharmaceutical companies. That is not generic customer research. It is work at the level of each workflow step.
Lesson Two: Compliance Can Be More Valuable than Model Quality
In professional fields, AI capability is not automatically trusted. Standards are trusted.
Elicit invested in PRISMA 2020 compliance instead of only chasing better summaries. Pharmaceutical teams do not buy because the prose sounds nicer. They buy when the output can satisfy field norms.
That reverse ordering, meet the standard first and then optimize the experience, applies across regulated markets such as healthcare, legal, and finance.
Lesson Three: Evolve from Tool to Platform
Elicit’s API strategy is worth studying:
- Start with a single-user tool and acquire users quickly.
- Add team collaboration to increase retention.
- Add API and platform capabilities to open the enterprise market.
This is not a new path, but Elicit executes each stage cleanly. It did not rush into platform claims before the product had traction. It launched the API after the core workflow had standing.
06 Risks and Challenges to Watch
Elicit still has obvious weak points.
Upstream dependency: Elicit depends on third-party LLMs. Its blog discusses evaluating models such as Claude Opus 4.5, Gemini 3 Pro, and GPT-5. If model providers change pricing or limit access, Elicit’s gross margin can be pressured.
General AI catching up: GPT-5 and Claude 4-class models are improving quickly at research tasks. If general models eventually produce PRISMA-compliant systematic reviews, Elicit’s differentiation could narrow.
Unclear revenue scale: The product has 5 million users, but paid conversion and ARR are not public. If paid users concentrate in lower-price tiers, commercialization may be less strong than the top-line user number suggests.
The Power of Vertical AI
Elicit represents an AI startup pattern that is becoming more credible: do not pursue the broadest user base; pursue the deepest workflow.
In the AI era, the most valuable product may not be the most general AI. It may be the most professional workflow replacement.
For a builder, the lesson is not Elicit’s AI technology. It is the productization mindset: find a specific group of people, understand their complete workflow, identify the highest pain point, redesign the whole flow with AI, and commercialize step by step.
That matters far more than calling a product “AI-native.”
Note: The data in this article is drawn from Elicit’s website and official blog. The product analysis is based on public information. Elicit was founded in 2022 and is treated here as a native new product.