What Claude (Anthropic) with web search answered for patent-search services (startups) on 2026-10-06

Claude (Anthropic) with web search claude-opus-5 asked 2026-10-06

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 mentioned
   Google Patents - start here, always.
2. Google Patents mentioned
   Google Patents - start here, always.
3. UpCounsel mentioned
   (UpCounsel's roundup (https://upcounsel.com/patent-search-tools) puts it in the same top tier as the official office databases.)
4. Espacenet mentioned
   Espacenet (EPO) + WIPO PATENTSCOPE - free, and they catch what Google misses.
5. WIPO Patentscope mentioned
   Espacenet (EPO) + WIPO PATENTSCOPE - free, and they catch what Google misses.
6. PATENTSCOPE mentioned
   Espacenet (EPO) + WIPO PATENTSCOPE - free, and they catch what Google misses.
7. Lens.org mentioned
   Lens.org - free/cheap, and the best bridge to scientific literature.
8. IPRally mentioned
   IPRally - the paid tool I'd pick first if you need one.
9. Solve Intelligence compared
   (softwarefinder.com/legal/iprally (https://softwarefinder.com/legal/iprally), Solve Intelligence's comparison (https://www.solveintelligence.com/blog/post/prior-art-search-ai-tools))
10. PatSnap mentioned
    PatSnap - only if you need competitive/landscape intelligence, not just search.
11. PatSeer mentioned
    PatSeer or Questel Orbit Express - the mid-market step up.
12. Questel mentioned
    PatSeer or Questel Orbit Express - the mid-market step up.
13. Questel Orbit mentioned
    PatSeer or Questel Orbit Express - the mid-market step up.
14. Orbit mentioned
    PatSeer or Questel Orbit Express - the mid-market step up.
15. Orbit Express mentioned
    PatSeer or Questel Orbit Express - the mid-market step up.
16. Orbit Intelligence mentioned
    Both are more affordable and less sales-heavy than full Orbit Intelligence or Derwent Innovation, and the common advice is exactly this path: free tools early, PatSeer or Orbit Express once you're actually preparing to file.
17. Derwent mentioned
    Both are more affordable and less sales-heavy than full Orbit Intelligence or Derwent Innovation, and the common advice is exactly this path: free tools early, PatSeer or Orbit Express once you're actually preparing to file.
18. Derwent Innovation mentioned
    Both are more affordable and less sales-heavy than full Orbit Intelligence or Derwent Innovation, and the common advice is exactly this path: free tools early, PatSeer or Orbit Express once you're actually preparing to file.
19. Klemchuk mentioned
    the AIPLA survey median for search-with-opinion is around $2,000, and firms like Klemchuk (https://www.klemchuk.com/flat-fee-patent-search) publish flat fees.
20. Freedom-to-operate mentioned
    Freedom-to-operate - budget for this separately, and scope it hard.

Sources the engine cited

The answer, as recorded

## What I'd actually use, in order **1. Google Patents - start here, always.** Free, full-text across 100+ patent offices plus non-patent literature, and its semantic "similar documents" feature is genuinely good. For the first question every startup has ("has someone already done this?"), it answers 80% of it in an afternoon. Zero reason to pay for anything before you've exhausted this. ([UpCounsel's roundup](https://upcounsel.com/patent-search-tools) puts it in the same top tier as the official office databases.) **2. Espacenet (EPO) + WIPO PATENTSCOPE - free, and they catch what Google misses.** Espacenet's classification browser (CPC) is the best free way to find the *right* technology bucket rather than just keyword-matching; PATENTSCOPE covers PCT applications and has decent machine translation for CN/JP/KR families. Use both as a cross-check on Google Patents before you conclude anything is novel. **3. USPTO Patent Public Search - free, for the one thing it's uniquely good at.** If you're filing in the US, this is the authoritative source for file histories and examiner-cited art. Reading the prosecution history of the two or three closest patents to your idea teaches you more about your odds than any analytics dashboard. **4. Lens.org - free/cheap, and the best bridge to scientific literature.** If your invention sits near academic research (biotech, materials, ML), prior art often isn't a patent at all - it's a paper or an arXiv preprint. Lens links patents to scholarly works in a way the office databases don't. **5. IPRally - the paid tool I'd pick first if you need one.** Graph-based AI rather than keyword or embedding soup, and critically it *shows you the graph* so you can see why it surfaced a result and refine it. That transparency matters when you're a non-expert trying to judge whether a hit is actually close. ISO 27001 certified, which your counsel will ask about. ([softwarefinder.com/legal/iprally](https://softwarefinder.com/legal/iprally), [Solve Intelligence's comparison](https://www.solveintelligence.com/blog/post/prior-art-search-ai-tools)) **6. PatSnap - only if you need competitive/landscape intelligence, not just search.** Broad global coverage (US/EP/JP/KR/CN) with analytics, citation maps, and export. It's the one startups most often buy, and often the one they most often regret: enterprise pricing, modular add-ons, tier limits, nothing published. Worth it if you're raising on an IP story or mapping a competitor's portfolio. Overkill for "can I file this?" ([Patsnap pricing on G2](https://www.g2.com/products/patsnap-analytics/pricing)) **7. PatSeer or Questel Orbit Express - the mid-market step up.** Both are more affordable and less sales-heavy than full Orbit Intelligence or Derwent Innovation, and the common advice is exactly this path: free tools early, PatSeer or Orbit Express once you're actually preparing to file. ([PatSeer pricing](https://www.g2.com/products/patseer/pricing), [Questel overview](https://us.fitgap.com/products/023804/questel)) **8. A flat-fee patentability search with written opinion - the one I'd actually pay for.** Honestly, this beats buying a software seat for most startups. A defined-scope US patentability search with opinion runs roughly $500-$1,500 for basic work and $2,000-$3,000 in a crowded field; the AIPLA survey median for search-with-opinion is around $2,000, and firms like [Klemchuk](https://www.klemchuk.com/flat-fee-patent-search) publish flat fees. You get a professional searcher *and* a legal read on what the art means - which is the part no tool gives you. ([UpCounsel on patent search cost](https://www.upcounsel.com/patent-search-cost)) **9. Freedom-to-operate - budget for this separately, and scope it hard.** FTO is a different and much more expensive animal: commonly $10,000-$50,000, though a tightly defined scope cuts that 30-50%. Staged pricing is normal - roughly $3,000 to find relevant live patents, $2,000-$4,000 to analyze them, and a non-infringement opinion letter starting near $6,000. Don't do FTO until you have a shipping product and a real infringement risk. ([LeanLaw on capping FTO costs](https://www.leanlaw.co/blog/how-to-structure-freedom-to-operate-fto-opinions-capping-costs-while-defining-clear-search-scope/), [Patent Trademark Blog](https://www.patenttrademarkblog.com/?p=647)) ## The honest version Most startups asking this question don't need a patent search *service* - they need about six hours on Google Patents and Espacenet, then a $1,500 flat-fee search with opinion before they spend real money filing. Buying a PatSnap or Orbit seat at seed stage is usually a way to feel like you're doing IP work rather than doing it. Two caveats worth naming: free tools are fine for novelty checks but genuinely insufficient for FTO or invalidity work, where missing one reference is the whole point. And much of the "top 10 patent search tools" content I found is written by the vendors themselves - PatSnap and Solve Intelligence both publish rankings that feature PatSnap and Solve Intelligence - so treat the comparative claims accordingly. The free office databases and the flat-fee-search recommendation are the parts I'd stand behind independently.

Near this answer

The same question, recorded on other dates: 2026-10-04, 2026-10-02, 2026-10-01, 2026-09-26, 2026-09-24, 2026-09-22, 2026-09-20, 2026-09-16, 2026-09-14, 2026-09-12, 2026-09-10.

patent-search services, asked for other audiences: small businesses (asked 2026-10-06).

Other categories with a recorded answer: accounting firms, business coaching services, business formation services, compliance consulting firms.

Markdown edition · JSON record · 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.