Webinar Q&A follow-up: your questions about IPRally’s search agents answered

Thanks to everyone who joined “Patent search agents in practice” on 18 November. You asked more than we could answer live, so here are the answers from our hosts John Paul Keeler, Yashasvi Ranawat and Luis Limon. They cover how to steer the Novelty and Invalidity Agents, what to give them, how to read the results, and where the FTO Agent stands. Missed a session? The recordings are linked at the end.

Working with the agents

Q: The Novelty Agent sometimes moves on before I have approved the claims and features. How do I step in?

A: Before you launch, open Options and turn on “Review claims before agent continues” and “Review features before agent continues”. The run then pauses at each step and waits for your approval. At the feature step you can edit, exclude or add features, and revert any change. After a run, you can also challenge a single analysis in the feature chart with the pencil icon. The agent re-reviews that document and explains its answer, citing the paragraphs it relies on.

Q: If I add my own claim set to the invention disclosure, does the agent use it?

A: Yes, as closely as possible, and it adds relevant information from the other materials you upload. Without a claim set, the agent writes the claims itself from the disclosure. IPRally’s search model is trained on claim-like language and performs best with it, and splitting the invention into independent and dependent claims lets the agent run a separate search along each one.

Q: Can I edit the features and re-analyse the documents already in the feature chart, without running the searches again?

A: Not as a separate step. In a Novelty run, editing the features and clicking Re-run regenerates everything from the features onwards, including the searches and the feature chart. Documents you have added yourself in collections are kept. If you only disagree with how one document was read, challenge that analysis instead of re-running. John Paul’s advice on features: keep each one to a single idea. The agent does better with more, narrower features.

Q: Why can I re-run the Novelty Agent but not the Invalidity Agent?

A: In the Invalidity Agent, the search target is the patent’s own claims, so there is nothing to regenerate. To change the features, start a new run with “Review features before agent runs” turned on, and edit them before the search starts. Within a run, you can still challenge analyses, override relevance tiers and add your own documents.

What to give the agent

Q: Does the Novelty Agent work better with claims only, or with the description and drawings as well?

A: Focused input works best. “The more noise you add, the more noise you’re going to get in your results,” John Paul says. The agent turns what you give it into claims, so a concise technical description of the inventive features, or the claims themselves, gives the cleanest result. To test a filed patent, paste its claims into a new Novelty run and set a date filter, rather than attaching the whole document. Drawings help when they show the inventive features: in the demo, the agent took the flexible striker flap of a snow thrower straight from the drawing. Use the context box to say what each image should focus on.

Q: Can the agent read 3D files, such as STEP or 3D PDF?

A: Not today. The agents accept text, PDFs, images and Office files. For a mechanical invention, export the key views as images and describe the important features in the context box. We have passed the request on to the product team.

Q: Can I add non-patent literature, such as articles from Google Scholar?

A: The agents search and analyse patent literature. You can upload a paper as input when you start a run, and the agent builds its claims and features from it, but it does not search or analyse NPL as prior art. To combine agent results with NPL, use the IPRally MCP. Copy the MCP prompt from an agent run into your own AI assistant, such as Claude or Copilot, and the assistant can read the agent’s project and report and set them against your NPL and internal sources. You can also take single collections, such as the examiner citations, out through the MCP.

Chemistry and life sciences

Q: Is IPRally useful for chemical structure searches, Markush claims or biological sequences?

A: IPRally searches patent text, not chemical structure databases. It has no substructure or superstructure search and no BLAST-type sequence search. Customers use the agents successfully with Markush structures by entering the structure, defining the R groups, and describing what the structure does and what is new about it. The agent then finds documents that disclose the same thing. “Your free text is your friend,” John Paul says.

Q: Can the agent recognise a molecule from an image? Is there a drawing tool?

A: The agent can identify a molecule from an image and include it in the claims, and the search can use it too. There is no drawing tool, so upload an image of the structure and add a description.

Dates, jurisdictions and search scope

Q: Why did the demo apply a publication date filter to a novelty search?

A: A new invention does not need one. The filter lets you rerun an older novelty case and compare the agent’s results with your original search, which is a good way to evaluate the agent on your own cases.

Q: Can I limit the Invalidity Agent to documents published before a cut-off date?

A: Yes. Every Invalidity run applies a cut-off: the priority date by default, the application date, or a custom date. The agent still shows intermediate documents, filed before the cut-off and published after it, because they can matter for novelty. A column in the feature chart shows which side of the cut-off each document falls on.

Q: For an EP patent, the Invalidity Agent returns HK or BR documents published after the cut-off, whose family members might be usable. Do I check these manually?

A: For now, yes. The search is global and there is no jurisdiction filter. The agent shows the closest match by AI score, and results are grouped by family, so the document shown may not be the family member you would cite. Check the family of any hit published after the cut-off. John Paul asked for feedback like this to come in through the Help Centre chat.

Q: Can the agents search only the title, abstract and claims?

A: Not as an agent setting. The agent reads full documents, because novelty-destroying disclosure can sit anywhere in the text. For a title, abstract and claims search, run it yourself in the agent’s project with a Boolean restriction, then add it under Document Sources. The agent analyses those results against the same features.

Batches and results

Q: Is there a quality difference between batches? Should I analyse every batch before looking at the feature chart?

A: Batches are ordered by AI score. The first batch holds the most promising documents, and each later batch the next most relevant. Review the first batch, then decide whether to analyse the next one, add your own documents, or search for features that came back absent. “Once you find your needle, you have the opportunity to stop,” Yashasvi says. You can generate the written report at any point, and the agent never re-analyses a document it has already read.

Q: How does the agent find second documents for inventive step?

A: The agent runs a separate search for each claim and claim combination, so documents that cover only some features still enter the pool. In Invalidity runs, the pool also includes searches on the target’s family and its examiner citations, and when a document proves highly relevant, its own citations are added too. The written report supports the inventive step analysis with structured evidence: the problem-solution approach for European targets, and a §103 analysis with suggested secondary references for US targets. Where a case rests on inventive step, add your own targeted searches to the project, and the agent analyses those documents against the same features.

Q: I added a collection to an agent run. How do I remove only that collection?

A: Go to Document Sources in the agent run and delete the collection there. You can also select several items and delete them together.

Training and benchmarks

Q: IPRally is trained on examiner citations. Does that make finding the known citations weak evidence?

A: Examiner citations are part of how the search model learns, but nothing boosts them in the results. Documents rank on AI score alone, and in the demo the three cited documents surfaced through normal recall. The model is trained on patent documents and examiner citations. It does not use litigation or PTAB decisions, office actions or file wrapper data, which avoids hindsight bias from cases that have already been decided. That also answers a related question: the reasoning in examiners’ written opinions is not part of training.

Q: Have you benchmarked the FTO Agent against manual analysis?

A: Not yet. The FTO Agent is still in development, and evaluation is part of the beta. The prototype shown in the webinar looks at in-force patents, pending applications and applications lapsed for unpaid fees. The beta is by invitation. Customers can register interest through their account manager, and others can contact sales@iprally.com.

What’s next

The Intelligence Agent is next. Join us on 20 January at 10:00 CET for a deep dive. Register here.

Missed a session in the Agents webinar? Watch the recordings here. To try the agents on your own cases, book a demo, or contact your account manager if you are already a customer.

‍

Read more

Webinar Q&A follow-up: your questions about IPRally’s search agents answered

Thanks to everyone who joined “Patent search agents in practice” on 18 November. You asked more than we could answer live, so here are the answers from our hosts John Paul Keeler, Yashasvi Ranawat and Luis Limon. They cover how to steer the Novelty and Invalidity Agents, what to give them, how to read the results, and where the FTO Agent stands. Missed a session? The recordings are linked at the end.

Webinar Q&A follow-up: your questions about the IPRally MCP answered

Thank you to everyone who joined Patent answers, inside the AI you already use. Across the two sessions on 4 November, you sent us more questions than we had time to answer live. We have grouped them by theme below, merged similar questions and included the answers given during the sessions.

Your IPRally workspace, wherever you work

Your AI assistant can now pick up where your IPRally work left off. The IPRally MCP now reads your collections, tags, search cases, projects, and agent runs, so your agent starts from results you’ve already vetted instead of a blank prompt.

Webinar Q&A follow-up: your questions about IPRally Drafting answered

Thank you to everyone who joined No more drafting in the dark: prior-art-aware patent drafting with IPRally Protection. Across our EU and US sessions, you sent us more than 80 questions, from file formats and drawings to claim clarity and US disclosure duties. We have grouped the answers by theme and merged similar questions. Some were answered live or in the chat during the sessions, and we include those here too.

Webinar Q&A follow-up: your questions about IPRally Intelligence answered

Your questions on IPRally Intelligence, answered: from working with larger datasets to verifying results at paragraph level.

IPRally is SOC 2 Type II compliant

IPRally has completed its SOC 2 Type II examination. Advantage Partners, our independent auditor, tested our security controls in operation from March 4 to June 4, 2026, and issued an unqualified opinion: no exceptions noted on any control tested.

When systems start talking: IPRally's journey toward a virtual IP department

Something new is happening: people talk to systems in plain language, systems talk back, and systems have started talking to each other the same way. As AI climbs from the small decisions to the large ones, the question for IP work is no longer whether it gets automated, but how to stay in control once it is. Co-founder and CPO, Sakari Arvela, lays out where this leads – through IPRally's new API and upcoming MCP support toward a virtual IP department that's always on – and why humans end up more important in that picture, not less.

Search, Intelligence, Protection: the three pillars of IPRally’s next chapter

For nearly a decade IPRally has been obsessed with getting one task right: patent search. Having built the best patent search in the world, we’re now reorganizing around three pillars – Search, Intelligence, and Protection – the newest of which makes drafting prior-art-aware from the first sentence. Co-founder and CPO Sakari Arvela maps out the next chapter, and why Protection is the shift he's most personally invested in.

NGen selects IPRally to bring AI-native patent intelligence to Canada's advanced manufacturers

After evaluating a range of tools on functionality, usability, and cost, Canada's Global Innovation Cluster for Advanced Manufacturing chose IPRally to power its Patent Search Adoption Tool Pilot Program.

IPRally and RPX Corporation announce strategic collaboration, cross-linking AI-powered patent search with RPX Empower

IPRally today announces a strategic collaboration with RPX Corporation, the global patent risk management service provider, that further expands the capabilities of the RPX Empower platform and provides seamless access to business critical complementary information for IPRally users.