How to complete a patentability search in IPRally in minutes using an image
Traditionally, patentability searches have been manual, slow, and dependent on deep expertise, but that's changing. In last week’s blog post, we showed how IPRally makes it possible to complete a professional patentability search in minutes using natural language input. This week, we’ll explore another approach: image-based searches. Start with any image and carry it through to review with IPRally’s AI assistants. In many cases, you’ll uncover documents that a non-visual search would have missed.

What is a patentability search?
A patentability search helps determine whether your invention is novel and eligible for patent protection. It supports R&D teams, patent professionals, and startups in assessing the risk of rejection and optimizing the scope of a patent application before filing. It also provides valuable insights into competitors’ activities in the invention’s field and helps evaluate its business potential.
Traditional searches are manual and heavily reliant on keywords, but IPRally’s AI automates the hard part: understanding the invention and comparing it with millions of documents.
IPRally interprets the technical meaning of your search input, identifies similar documents based on technical similarity rather than just matching words, and then performs an in-depth analysis with professional-level reasoning. In the example shown in the video in this blog post, we’ll use a picture of a snow thrower as input, but you can use almost any image – from photos and drawings to product sheets or molecule diagrams. You can of course use natural language text as input as well.
Why use images as search input?
Image search is fast, simple, and powerful – on its own or as a complement to text-based searches. It’s especially useful when a concept is difficult to describe in words, or if you're dealing with an unfamiliar technology field. By focusing on conceptual similarity rather than visual matches, Image search can uncover patents you might otherwise miss, providing a more complete view of the landscape. From technical drawings and schematics to product sheets and rough sketches, IPRally can interpret a wide range of visual inputs. Used alongside text-based searches, it helps ensure you don’t overlook relevant documents.
How to do it:
1. Upload an image of your invention concept
You can start with almost any type of image: a mechanical drawing, a photo, a chemical structure diagram, a hand drawn sketch or a product sheet. IPRally can read and understand all of them, including any text within the image. Unlike a typical image search engine, IPRally does not look for visually similar images. Instead, its Graph AI interprets the technology shown in your image, then finds documents describing similar concepts, even if they appear completely different visually or contain no images at all.
2. Provide context if needed
Sometimes an image alone is not enough to generate an accurate search. The image might be incomplete or easily mistaken for something else – for example, a rough drawing of a copper wire could resemble an optical cable. In these cases, adding a few words in the Context field will guide the AI in the right direction. Simply describing what’s pictured and its intended use is usually sufficient. If you are using a patent drawing, defining the numbered parts or their properties can lead to even more precise results.
3. Run the search and browse results
Click “Search Patents” and IPRally will interpret the technical concept of your image (and any context) and create a knowledge graph, which is a tree-like structure of the invention's key features, their functions and their relations.
IPRally then compares this knowledge graph against millions of patents to identify the most technically similar documents – independently of images, keywords, language, and official classifications.
4. Use IPRally’s AI assistants to filter and understand the results
Once you have run your image search, it’s time to make sense of the search results. IPRally gives you two powerful tools to explore and refine the list: Ask AI and Smart Filters.
With Ask AI, you can ask natural language questions across multiple documents without having to read them. Just apply it to all of the search results to get a quick start for your review. Here’s an example giving you immediate useful information for your review process:
“Does the invention mention an auger driven by a motor to rotate?”
If your question is binary like above, i.e. can be answered with a Yes or No, you can apply Smart Filters to filter out the irrelevant hits from the search results. In addition, you will get a full reasoning for the answer, grounded to specific paragraphs in each of the documents.
In our sample case, suitable binary questions could be for example:
"Is the auger house made of at least two different materials?"
"Does the invention mention the integration of the handle device for user operation and the frame?"
The results are automatically grouped based on AI’s answer: Yes, No or Maybe. Select “Hide No Patents” to instantly narrow down your results to the most relevant documents with no manual reading required. For further filtering, just open the Smart filters panel to filter on either Yes, No or Maybe.
After a few rounds with Ask AI and Smart Filters, you’ll likely narrow your results to a handful of highly relevant hits.
In just minutes, starting from an image, you have conducted a professional patentability search.
Watch the video below to see the process in action.
Pro tips for image-based patentability searches
Aim for clarity and detail: Use the clearest and most detailed images you can. Hi-resolution images or precise drawings help IPRally interpret the invention more accurately.
Sometimes, context is king: If the initial interpretation of the image is inaccurate, add a few words describing what’s pictured and its intended use. This is usually enough to guide the AI in the right direction. If a drawing only shows reference numbers without labels, you can provide the meanings in the additional context field (e.g., "1: Auger housing, 2: Motor, 3: Handle").
Run multiple searches: Using different images of the same invention saturates the search with information and improves results. IPRally's Projects feature makes it easy to manage multiple searches, along with their Ask AI history and comments, in one dedicated environment.
Try mixed searches: Text-based and image-based searches often yield overlapping but slightly different results, so combining them can give you greater confidence in the completeness of your search. Simply open a new Project and use both approaches together.
Fine tune your search: After using Ask AI and Smart Filters, you’ll likely have a few strong hits marked as Favorites. With Zoom to Favorites, you can trigger a new AI search focused on the technical features of those documents. This helps uncover similar patents that may have been overlooked in the initial search. You can also edit the knowledge graph and apply Boolean filters to refine your search.
Trust data security: All information entered into IPRally is encrypted both in storage and in transit, and IPRally employees cannot access your input. Images are not used for training AI models, and no customer-identifying information is shared with third parties.
Get started with your own patentability search
Want to see how easy and accurate an image-based patentability search can be?
- Drag and drop an image of your invention
- Let IPRally’s AI do the heavy lifting
- Get novelty-relevant prior art in minutes – no Boolean required
Start a free trial – All features mentioned in this guide are included
Get a live demo – See invalidity workflows tailored to your specific needs