Generative AI has finally made a serious entry into the realm of patent search. Perplexity Patents, released on October 30 by Perplexity, is a tool that allows anyone to simply ask natural-language questions such as “What are the key quantum computing patents after 2024?” or “Are AI systems for language learning patented?” and instantly receive a list of relevant patents—with direct links to the original publications. It’s free during the beta period, and users of Pro/Max plans will gain access to additional searches and options. This is the kind of service that truly embodies the idea of “a world where everyone can access patents.”
Below, we’ll unpack this development from three perspectives: Japanese inventors, in-house IP teams, and patent firms.
What’s new about it?
Until now, one could already do quite a lot by combining tools like Google Patents, J-PlatPat, and Espacenet. However, all of them required users to manually craft search queries, understand FI/F-terms or CPC classifications, and handle keyword variations (AI / artificial intelligence / machine learning / deep learning, etc.) on their own. Perplexity Patents adds a new layer — one where “you can just ask in natural language, and the AI expands your query intelligently.” It’s similar to how asking about “fitness trackers” also brings up adjacent technologies like smart bands or healthcare wearables.
What’s even more interesting is that it doesn’t limit prior art to patent publications. Perplexity’s core strength has always been aggregating the web, academic papers, code, and open repositories — and it extends that to the non-patent literature (NPL) side of prior art. The team states that it searches blogs, videos, and open-source software as well, which means it can surface gray-area disclosures—cases where an invention was made public but never patented. For Japanese practitioners, that’s a particularly relevant feature.
Why this doesn’t yet make prior-art searches “complete”
In practice, two major concerns remain: coverage and accountability.
- Coverage transparency
Perplexity claims to maintain a global knowledge index of patent data, but it’s unclear how up-to-date and geographically comprehensive that database is. While US, EP, and PCT publications may be well covered, the update frequency for Japanese or Chinese publications—especially the linkage between laid-open and granted patents—remains opaque. For corporate freedom-to-operate (FTO) investigations, that opacity is critical. As long as this remains a black box, you can’t justify a report that says, “We asked the AI, and it found nothing.”
- Search intent visibility
Natural-language querying is convenient, but without knowing which IPC/CPC codes or keywords the AI actually used, users can’t verify completeness. Traditional search reports describe “which F-term groups and keywords were used” to ensure reproducibility in examination or litigation. Unless AI-generated queries are similarly auditable, they can’t yet take a central role in legal work.
- Hallucination and NPL reliability
Perplexity mitigates the risk of fabricated citations by always providing direct links, unlike ChatGPT’s freeform text. Still, it may miss NPL items or overinclude noisy data. In the end, human searchers are still needed to refine and validate the results.
Why this is most impactful for Japanese inventors and researchers
The greatest benefit is in early-stage idea screening. At the concept stage, inventors can now ask “Is this field already saturated in the US?” and get a rough answer in 30 seconds. That makes it easier to identify differentiation points when drafting invention disclosure forms. What previously required asking the IP department can now be done independently, improving the quality of disclosures and reducing redundant filings.
The ability to query in ambiguous Japanese is also significant. The biggest hurdle with J-PlatPat is often figuring out the first keyword to enter. AI that can interpret those vague queries will greatly lower the entry barrier. In fact, Japanese users are already experimenting to see whether Perplexity can “search patents in natural language” and “display citation counts,” suggesting that professionals will soon start testing it hands-on.
The right stance for patent firms and corporate IP teams
This should be seen not as “replacement,” but as democratization of the front-end:
- Let inventors and R&D teams use Perplexity Patents freely.
- Have the IP department recheck and re-evaluate only the “interesting” results identified by AI through official search channels.
- Create internal guidelines explaining how the AI picked up those results.
Instead of jumping straight into using it for FTO or invalidation searches, the safer starting point is to apply it to idea generation and competitive landscape scanning.
For patent firms, this also opens an opportunity to create educational content: for instance, an article or seminar comparing “what you can see with Perplexity Patents” versus “what’s required in professional practice.” It’s a perfect moment to demonstrate where the human expertise of patent attorneys remains indispensable in the AI era.
Likely developments ahead
Perplexity has hinted at a scholarly version, Perplexity Scholar, aimed at academia. When research papers and patents can be viewed side by side, cross-checking will become seamless: researchers can see that “the method appeared in a paper first but hasn’t been patented yet,” or conversely, “the patent came first, so publishing a paper might be risky.” That integration could make corporate R&D–IP collaboration far more real-time. By around 2026, we may see Japanese organizations designing workflows where “AI drafts invention disclosures while simultaneously retrieving similar patents.”
Conclusion
Perplexity Patents is a tool that begins to dismantle the era when patent search was confined to specialists. But it only disrupts the entry point—the exit points of patent analysis (ensuring completeness, generating evidence, interpreting claims, constructing invalidation grounds) still firmly belong to humans and traditional databases.
The real challenge now is to rethink how invention creation and internal communication evolve in a world where the entry barrier is gone. AI may now find prior art for you, but it won’t write your technical distinctions or embodiments. In fact, the more “similar prior art” AI surfaces, the clearer and more differentiated your patent specification must become—and that’s exactly where the skill of a true practitioner will shine.