What Claude (Anthropic) with web search answered for patent-search services (startups) on 2026-09-07

Claude (Anthropic) with web search claude-opus-5 asked 2026-09-07

Coverage: Claude (Anthropic) with web search. Other engines not included.

The question asked, verbatim: best patent-search services for startups

Brands named, in the order the engine gave

1. AI mentioned
   The $30k/year AI platforms are almost never worth it pre-Series B.
2. Google Patents mentioned
   - Google Patents - the best starting point.
3. Patents mentioned
   - Google Patents - the best starting point.
4. PQAI mentioned
   (PQAI (https://projectpq.ai/5-free-and-possibly-more-efficient-alternatives-to-google-patents/))
5. Espacenet mentioned
   - Espacenet (EPO) - ~150M documents, the most comprehensive free source, with CPC classification, citation and patent-family search plus machine translation from 30+ languages.
6. USPTO Trademark Search mentioned
   - USPTO Patent Public Search - official US source;
7. Lens.org mentioned
   - Lens.org - free, and uniquely good at linking patents to scholarly literature, which matters if you're in biotech/deep tech where the killer prior art is a paper, not a patent.
8. Good Grants mentioned
   Good complement to keyword search;
9. Cardinal Intellectual Property mentioned
   - Cardinal Intellectual Property - large, well-established US firm;
10. Ensemble IP mentioned
    - Ensemble IP - the ex-Landon IP team, searching since 2004;
11. Clarivate mentioned
    - Clarivate (Landon IP) - same lineage, now inside Clarivate;
12. GreyB / Sagacious IP mentioned
    - GreyB / Sagacious IP - India-based, very competitive pricing and fast turnaround;
13. Sagacious IP mentioned
    - GreyB / Sagacious IP - India-based, very competitive pricing and fast turnaround;
14. IPRally mentioned
    - IPRally - graph-based AI trained on patent text;
15. Patlytics mentioned
    - Patlytics - broadest lifecycle coverage (search, analysis, infringement charting);
16. Solve Intelligence mentioned
    - Solve Intelligence - drafting-first with search attached;
17. XLSCOUT mentioned
    - XLSCOUT - strongest for landscape/scouting rather than single-invention checks.
18. PatSnap mentioned
    PatSnap ($30k-$60k/yr) and Questel Orbit ($25k-$50k/yr) are excellent analytics platforms with deep EP/Asia coverage, but they're priced for corporate IP departments, not seed-stage companies.
19. Questel Orbit mentioned
    PatSnap ($30k-$60k/yr) and Questel Orbit ($25k-$50k/yr) are excellent analytics platforms with deep EP/Asia coverage, but they're priced for corporate IP departments, not seed-stage companies.

Sources the engine cited

The answer, as recorded

Short version: for most startups the right answer is a free database for your own first pass, then a few hundred dollars to a professional search firm before you spend real money on filing. The $30k/year AI platforms are almost never worth it pre-Series B. ## 1. Do it yourself first (free) - **Google Patents** - the best starting point. Full-text across 100+ offices, machine translations, "similar documents," and it indexes non-patent literature (papers, some books). Fastest way to find out whether your idea is obviously already patented. Weaker on classification-based searching and non-English filings than the official databases. ([PQAI](https://projectpq.ai/5-free-and-possibly-more-efficient-alternatives-to-google-patents/)) - **Espacenet** (EPO) - ~150M documents, the most comprehensive free source, with CPC classification, citation and patent-family search plus machine translation from 30+ languages. This is where you catch the CN/JP/KR/DE art Google surfaces poorly. ([Espacenet](https://worldwide.espacenet.com/), [PatSeer](https://www.patseer.com/free-and-paid-patent-databases-2026-complete-guide/)) - **USPTO Patent Public Search** - official US source; proper Boolean and fielded search, plus file wrappers, assignments and legal status. Use it when you need to be sure about a specific US reference. ([USPTO](https://www.uspto.gov/patents/search/patent-public-search)) - **Lens.org** - free, and uniquely good at linking patents to scholarly literature, which matters if you're in biotech/deep tech where the killer prior art is a paper, not a patent. ([Cypris](https://www.cypris.ai/insights/patsnap-alternatives-8-tools-for-patent-research-intelligence)) - **PQAI** - free AI/natural-language prior art search: paste your invention description in plain English instead of guessing keywords. Good complement to keyword search; there's a $199/mo API tier if you want to automate it. ([PQAI](https://projectpq.ai/top-patent-search-apis-for-enterprises/)) ## 2. Then pay a human searcher (the highest-value spend) A professional patentability search typically runs a few hundred to ~$2,000 depending on complexity - cheap insurance against a $10-15k filing on unpatentable subject matter. ([Menteso](https://menteso.com/blog/how-much-the-prior-art-search-cost/)) - **Cardinal Intellectual Property** - large, well-established US firm; does patentability, clearance, landscape and accelerated-exam searches. Solid default. ([Cardinal](https://cardinal-ip.com/patent-search/)) - **Ensemble IP** - the ex-Landon IP team, searching since 2004; strong on international/foreign-language art. ([Ensemble IP](https://ensembleip.com/)) - **Clarivate (Landon IP)** - same lineage, now inside Clarivate; use if you want it bundled with Derwent data. ([Clarivate](https://clarivate.com/intellectual-property/landon-ip-is-clarivate/)) - **GreyB / Sagacious IP** - India-based, very competitive pricing and fast turnaround; common choice for startups on a budget. ([Sagacious](https://sagaciousresearch.com/prior-art-search)) One caveat: if the question is *freedom to operate* rather than patentability - i.e., "can I ship this without being sued" - order it **through your patent attorney**, not directly. Same searchers, but the analysis stays under privilege. ## 3. AI platforms - only if you're searching continuously - **IPRally** - graph-based AI trained on patent text; the most credible of the AI-native search tools for patentability and invalidity work. Roughly $24k-$72k/year. ([Software Finder](https://softwarefinder.com/legal/iprally)) - **Patlytics** - broadest lifecycle coverage (search, analysis, infringement charting); $21M raised, Google and Quinn Emanuel as customers. ([Patentext](https://patentext.com/blog/solve-intelligence-vs-patlytics/)) - **Solve Intelligence** - drafting-first with search attached; $40M Series B in Dec 2025. Worth a look if you're generating a lot of applications in-house. ([Patentext](https://patentext.com/blog/solve-intelligence-vs-patlytics/)) - **XLSCOUT** - strongest for landscape/scouting rather than single-invention checks. ([G2](https://g2.com/products/xlscout/pricing)) ## 4. Enterprise databases - skip for now **PatSnap** ($30k-$60k/yr) and **Questel Orbit** ($25k-$50k/yr) are excellent analytics platforms with deep EP/Asia coverage, but they're priced for corporate IP departments, not seed-stage companies. Revisit when you have a portfolio to manage rather than a first filing to make. ([Wicely](https://wicely.com/resources/patent-monitoring-platform-comparison)) **What I'd actually do:** an afternoon in Google Patents + Espacenet + PQAI to kill obviously-dead ideas, then ~$500-$1,500 to Cardinal, Ensemble, or GreyB on the one or two ideas that survive, then talk to a patent attorney. Total under $2k before you commit to filing.

Markdown edition · the question page, with every recording · Answer Watch · AEO Watch

Answer Watch records what one AI answer engine said on one date, nothing more. Each page names the engine and the model, carries the date the question was asked, and lists the brands that appeared in the answer in the order the engine gave them, with the sources it cited. It is a measurement, not a ranking, not a recommendation and not a claim about the world. Brands that did not appear in an answer are not named on any public page. Coverage: Claude (Anthropic) with web search. Other engines not included.