What Claude (Anthropic) with web search answered for patent-search services (startups) on 2026-09-16

Claude (Anthropic) with web search claude-opus-5 asked 2026-09-16

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. Google Patents mentioned
   Google Patents - The first place I'd send anyone.
2. Espacenet mentioned
   Espacenet (EPO) - Use it alongside Google Patents, not instead of.
3. Lens.org mentioned
   Lens.org - Largest single index (~158M patent documents), links patents to scholarly literature, and the free tier is generous.
4. PQAI mentioned
   AI-assisted search tools - PQAI (free) and PatentScan/Cypris-class tools - PQAI lets you describe your invention in plain English and returns conceptually similar art, which is exactly the failure mode of keyword search.
5. Cypris mentioned
   AI-assisted search tools - PQAI (free) and PatentScan/Cypris-class tools - PQAI lets you describe your invention in plain English and returns conceptually similar art, which is exactly the failure mode of keyword search.
6. PatSnap mentioned
   PatSnap, Derwent Innovation (Clarivate), Orbit Intelligence (Questel) - These are the serious professional platforms, and they're excellent: Derwent has the deepest normalized data and best citation mapping, Orbit is strongest on freedom-to-operate and European coverage, PatSnap has the best visuali
7. Derwent mentioned
   PatSnap, Derwent Innovation (Clarivate), Orbit Intelligence (Questel) - These are the serious professional platforms, and they're excellent: Derwent has the deepest normalized data and best citation mapping, Orbit is strongest on freedom-to-operate and European coverage, PatSnap has the best visuali
8. Derwent Innovation mentioned
   PatSnap, Derwent Innovation (Clarivate), Orbit Intelligence (Questel) - These are the serious professional platforms, and they're excellent: Derwent has the deepest normalized data and best citation mapping, Orbit is strongest on freedom-to-operate and European coverage, PatSnap has the best visuali
9. Clarivate mentioned
   PatSnap, Derwent Innovation (Clarivate), Orbit Intelligence (Questel) - These are the serious professional platforms, and they're excellent: Derwent has the deepest normalized data and best citation mapping, Orbit is strongest on freedom-to-operate and European coverage, PatSnap has the best visuali
10. Orbit mentioned
    PatSnap, Derwent Innovation (Clarivate), Orbit Intelligence (Questel) - These are the serious professional platforms, and they're excellent: Derwent has the deepest normalized data and best citation mapping, Orbit is strongest on freedom-to-operate and European coverage, PatSnap has the best visuali
11. Questel mentioned
    PatSnap, Derwent Innovation (Clarivate), Orbit Intelligence (Questel) - These are the serious professional platforms, and they're excellent: Derwent has the deepest normalized data and best citation mapping, Orbit is strongest on freedom-to-operate and European coverage, PatSnap has the best visuali
12. Questel Orbit mentioned
    - Questel Orbit Alternatives: AI Patent Search Tools Compared (2026) (https://www.patentscan.ai/blog/questel-patent-search-orbit-alternatives-guide-5go7)

Sources the engine cited

The answer, as recorded

Here's how I'd actually rank these for a startup, roughly in the order I'd reach for them: ## Start free - this covers most startups' real needs **1. Google Patents** - The first place I'd send anyone. Full-text search across 100+ patent offices, non-patent literature (papers, books), automatic machine translation of foreign-language patents, and the "similar documents" feature that surfaces prior art you'd never find by keyword. The 2026 AI Pro tier adds conversational search and image-to-patent matching at no cost. For a founder doing a first-pass knockout search before spending money on counsel, this is genuinely enough. **2. Espacenet (EPO)** - Use it alongside Google Patents, not instead of. Covers 90+ countries and is the best free tool for patent *families* - seeing every jurisdiction where a given invention was filed. Its classification (CPC) browsing is better than Google's, and CPC-based searching is how you find prior art that uses different vocabulary than yours. **3. USPTO Patent Public Search** - This is the actual tool USPTO examiners use (PE2E). If you want to see what the examiner reviewing your application will see, including pre-grant publications and OCR'd pre-1970 patents, search here. Clunky interface, but authoritative for US filings. **4. Lens.org** - Largest single index (~158M patent documents), links patents to scholarly literature, and the free tier is generous. Good for landscape views - who's filing in your space - rather than pinpoint novelty checks. ## When you need something defensible **5. A dedicated search firm - Cardinal IP is the one most patent firms quietly outsource to.** Around $800 for a professional search with 150+ technical researchers across all fields. Your law firm will charge you ~$3,000 for a search-plus-opinion package and send roughly $800 of that to Cardinal. Going direct saves real money if you already have counsel who'll write the opinion. Budget $1,500-$3,000 for the full attorney search-plus-patentability-opinion if you want the legal conclusion, not just the references. **6. AI-assisted search tools - PQAI (free) and PatentScan/Cypris-class tools** - PQAI lets you describe your invention in plain English and returns conceptually similar art, which is exactly the failure mode of keyword search. Free tier caps at 20 results with no full text or export, so treat it as a screening layer, not a replacement. ## Skip these (for now) **PatSnap, Derwent Innovation (Clarivate), Orbit Intelligence (Questel)** - These are the serious professional platforms, and they're excellent: Derwent has the deepest normalized data and best citation mapping, Orbit is strongest on freedom-to-operate and European coverage, PatSnap has the best visualization and the friendliest UI. But reported pricing runs $25K-$80K/year, and Derwent reportedly takes 40-80 hours of training to use properly. Not a startup purchase unless IP *is* your business model - a patent-heavy biotech or a licensing play. Note that much of the comparison data here comes from vendor-adjacent blogs, including PatSnap's own, so treat specific numbers as directional. **Two things worth flagging:** First, a search and a patentability opinion are different products. Free tools tell you what exists; only an attorney gives you an opinion you can rely on, and only the opinion has legal weight. Second, if you're doing a **freedom-to-operate** search - "can I ship this without getting sued?" - rather than a novelty search, do not DIY it. FTO turns on claim construction and in-force status, and getting it wrong is how you end up with willful infringement exposure. That's a lawyer job from the start. Sources: - [Top 10 AI Patent Search Tools For Novelty and Prior Art Search - PQAI](https://projectpq.ai/top-ai-patent-search-tools-2026/) - [Best AI Patent Search Tools in 2026 - Cypris](https://www.cypris.ai/insights/best-ai-patent-search-tools-in-2026-the-definitive-guide-for-r-d-and-innovation-teams) - [PatSnap vs Derwent vs Orbit Intelligence: The 12-Criteria Comparison](https://beyondelevation.com/blog/posts/patsnap-vs-derwent-vs-orbit-comparison/) - [Best 7 Patent Search Platforms - PatSnap](https://www.patsnap.com/resources/blog/articles/best-patent-search-platforms-2025/) - [Questel Orbit Alternatives: AI Patent Search Tools Compared (2026)](https://www.patentscan.ai/blog/questel-patent-search-orbit-alternatives-guide-5go7) - [Cardinal IP - Patent Search](https://cardinal-ip.com/patent-search/) - [Patentability Search Cost: What to Budget - CASRAI](https://casrai.org/guides/patentability-search-cost) - [Why Should You Conduct a Patent Search? - BitLaw](https://www.bitlaw.com/guidance/patent/why-conduct-patent-search.html) - [Patent Public Search - USPTO](https://www.uspto.gov/patents/search/patent-public-search) - [Espacenet](https://worldwide.espacenet.com/) - [Quick Guide to Free Patent Search - University of Maryland Libraries](https://lib.guides.umd.edu/patent_tm/patent_search_tips)

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.