If you are evaluating AI drug discovery platforms for a partnership or investment, the single most important fact is this: no standardized, cross-platform "preclinical success rate" exists to compare them on. A recent review of the field found there is no universally accepted set of key performance indicators for AI in drug discovery, and that metrics like time-to-IND and lead optimization success are rarely reported consistently. The practical alternative is to compare platforms on their business models, data strategies, and the observable structure of their pharmaceutical partnerships — which is what this article does.
The Challenge of Evaluating AI Drug Discovery: Beyond 'Success Rates'
The temptation to ask "which platform has the best success rate?" is understandable and unanswerable. The review published in PMC's journal archive identifies the core problem: outcome reporting in the field is often narrative or descriptive, which limits objective cross-study assessment and weakens the ability to measure domain-wide impact. When a company says its platform "accelerated" a program, there is rarely a comparable denominator behind that claim.
The clinical record so far reinforces the caution. A narrative review for medical professionals published by Noah AI notes that no AI-discovered drug has yet achieved regulatory approval, and that Phase II success rates for AI-discovered drugs align with historical industry averages of roughly 40%. The most clinically mature AI-discovered asset, rentosertib from Insilico Medicine, is still the exception rather than a trend.
Drug Discovery News draws the sharpest distinction I have seen: AI target identification has genuinely changed how many candidate hypotheses a team can generate and how quickly it can rank them, but it has not yet changed how many of those hypotheses turn out to be right once biology gets the final vote. When a platform pitches you, the critical question is which of those two things it is actually improving.
