Verified against ChatGPT · 2026-08-13
Draft a patent landscape brief that flags prior-art risk instead of quietly hiding it
Produces a preliminary patent landscape brief for internal use that names potentially conflicting prior art and open uncertainty explicitly, framed clearly as a research aid for an attorney's review rather than a legal opinion.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Draft a preliminary patent landscape brief for internal use — this is a research aid to help prepare for a conversation with a patent attorney, not a legal opinion, and the output must say so plainly.
INVENTION OR FEATURE BEING ASSESSED
A method for automatically re-ranking search results based on a user's real-time scroll behavior within the same session.
WHAT MAKES IT DIFFERENT FROM EXISTING APPROACHES, AS I UNDERSTAND IT
Most re-ranking systems we're aware of use click data after the fact; ours adjusts ranking live, mid-session, based on scroll dwell time before any click happens.
KNOWN SIMILAR PRODUCTS OR PATENTS I'M AWARE OF
A major search engine's public engineering blog post describing a session-based re-ranking signal, and one competitor's patent filing we found mentioned in a news article.
INTENDED USE OF THIS BRIEF
Prep material for a 30-minute call with outside patent counsel before deciding whether to file.
Open the brief with an explicit statement that this is a non-legal, preliminary research aid only, that it does not constitute legal advice or a formal freedom-to-operate opinion, and that any launch or filing decision requires review by a qualified patent attorney. Do not assert that any specific patent is or is not infringed, and do not state as fact whether A method for automatically re-ranking search results based on a user's real-time scroll behavior within the same session. is or isn't novel — instead, describe what Most re-ranking systems we're aware of use click data after the fact; ours adjusts ranking live, mid-session, based on scroll dwell time before any click happens. appears to be based on the information given, name the specific similar products or approaches in A major search engine's public engineering blog post describing a session-based re-ranking signal, and one competitor's patent filing we found mentioned in a news article. that look closest to it, and flag exactly where the overlap or distinction seems to lie, explicitly framed as "this looks worth having an attorney check" rather than a conclusion. Where you don't have enough information to assess overlap with a named prior approach, say so and note that as an open question for the attorney conversation rather than guessing. Never invent a specific patent number, filing date, or legal standard as if you had looked it up — if the brief needs a specific existing patent identified, instruct me to have that pulled from an actual patent database or provided by counsel rather than presenting an invented one.
WHAT NOT TO DO
Do not soften this into pure reassurance ("this looks clearly novel") or pure alarm ("this clearly infringes") — either overclaim is worse than the honest, bounded uncertainty this brief is supposed to represent.
OUTPUT FORMAT
1. The non-legal-advice disclaimer, stated plainly at the top.
2. A one-paragraph description of the claimed novelty as understood.
3. A table of known similar art with the specific point of overlap or distinction for each, and a confidence note (clear overlap worth flagging / unclear, needs attorney review / distinction looks meaningful based on information given).
4. A closing list of the specific open questions to bring to the attorney conversation.Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
This is deliberately scoped as a non-legal internal research aid, not a substitute for counsel, because a model has no reliable way to perform an actual freedom-to-operate search against the real patent record, and asserting novelty or infringement as fact would present an unverified guess with the confidence of a real legal opinion — the explicit disclaimer at the top, plus the instruction never to invent a specific patent number or filing date, exists precisely to prevent that guess from being mistaken for verified legal fact. The three-tier confidence framing (clear overlap worth flagging / unclear, needs review / distinction looks meaningful) matters mechanically because a binary yes/no on novelty forces the model to resolve genuine uncertainty in one direction or the other, and either resolution overclaims confidence it doesn't actually have; a graded confidence note lets the brief honestly represent "I can see why this might matter, but I can't determine the actual answer" as a legitimate output rather than forcing a false binary. Explicitly forbidding both pure reassurance and pure alarm addresses the two failure modes a model tends toward under ambiguity — either agreeably confirming the user's belief that their invention is novel (since that's what the framing implies they want to hear), or overcorrecting into exaggerated caution that treats every tangential similarity as a serious risk; both directions are less useful to an attorney than a plainly stated, specific list of open questions. The instruction to route any need for a real patent number to an actual database or counsel, rather than fabricating one, is the direct guardrail against the single most damaging failure mode here: a specific-sounding but invented patent citation would look exactly as credible as a real one to someone relying on this brief.
What you get back
This is a preliminary, non-legal research aid only — it does not constitute legal advice or a freedom-to-operate opinion, and any filing or launch decision should be reviewed by a qualified patent attorney before acting on it. Claimed novelty: live, pre-click scroll-based re-ranking versus post-click historical re-ranking. Known similar art: [Search Engine] blog post describing session-based signals — unclear, needs review, since the blog post doesn't specify whether it's pre-click or post-click. Open questions for counsel: does the described competitor patent filing cover pre-click behavioral signals specifically, or only post-click data.
Verified against
ChatGPT GPT-5.1 · 2026-08-13
Changelog
- 2026-08-13 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
