Patent Bots Documentation

Our products are mostly self-explanatory so you can use them without reading this documentation, but you'll find some helpful tips here.

We also have a recorded a demo that you can access below. The demo starts with PatentPlex, continues to patent proofreading at 2:40, and then to patent examiner statistics at 23:49.

Automated Proofreading

Our automated proofreading includes proofreading of patent applications and office action responses. Our proofreading tools are available on our website and also from a Microsoft Word add-in.

Patent Applications

We have two types of processing for patent applications: (1) "Analyze Claims" and (2) "Analyze All" which checks reference labels and figure numbers in addition to claims. We process your patent application in memory (never stored to a database) and immediately discard all information after returning results.

Note that our patent proofreading isn't perfect, and we do make mistakes. We will, however, find mistakes that you miss, and greatly improve the quality of your work. We believe that we have the most accurate patent proofreading available.

Expected Format

We are only able to process MS Word documents in the ".docx" format. If your file has a ".doc" extension, please save it as a .docx file.

We expect the following sequence in your patent application:

  • Title — We use the first non-empty paragraph as the title. If you have a header page, then we may have an incorrect title, but this won't affect our processing.
  • Specification — All paragraphs up to the beginning of the claims section are included in the specification.
  • Claims — We identify the start of your claims section by looking for a paragraph with only the word "Claims" or a phrase like "What is claimed is:" (any capitalization). All paragraphs until the abstract section are included in the claims.
  • Abstract — We identify the start of the abstract section by looking for (i) a paragraph "Abstract" or "Abstract of the Disclosure" (any capitalization) or (ii) a non-numbered paragraph after a page break.

We have tried to accomodate different styles of practice, but if your style differs from what we expect, you may get an error. If you do get an error, please contact us for assistance. We will do our best to update our processing to accomodate your practice.

A few items to note about how we process your documents:

  • the/said — If your claims use "said" as in "said widget", we will change "said" to "the". This improves our accuracy in processing your claims.
  • Comments — We allow you to include comments in the claims, e.g., to share with inventors. All paragraphs in the claims that start and end with "***" will be treated as comments and ignored.
  • Track Changes — Track changes are treated as if they had been accepted in your Word doc.

Analyze Claims

Our analysis results are self explanatory, but we will highlight a few features that you may not have noticed. For several parts of our analysis results, you can click or double click to obtain more information:

  • For the antecedent basis results, click on any yellow highlight, red highlight, or red underline to see additional information to help you diagnose the error. When you do this, we will bold relevant claim words to help you understand the error.
  • For the antecedent basis results, you can also hide part of the highlighting. For example, you may want to hide the green highlights. See the check boxes near the top of the tab.
  • For word and phrase support, hover over any word/phrase to the number of matches and double click to immediately see how you used the word/phrase in the spec.
  • For phrase support, you can also check support for any phrase in the claims. Select a phrase with your mouse, and we'll immediately count the number of times that phrase appears in the spec and show it to you in the floating box on the right. You can also double click the selected phrase in this box to see how you used the selected phrase in the spec.

We also support multiple dependent claims. For multiple dependent claims, to not have an antecedent basis error, you must have an antecedent basis in each of the parent claims. For example, if claim 20 recites "the widget" and depends from claims 1-19, then each of claims 1-19 (or one of their parents) must recite "a widget" to not have an antecedent basis error.

Analyze All

We check your reference labels and figure numbers for errors in addition to the claim analysis described above. We can do this with only your application, but we can also process your figures to do additional error checking. We can process figures in PowerPoint (must be .pptx and not .ppt), Visio (must be .vsdx and not .vsd), and PDF formats (must include text since we don't yet OCR).

We look for reference labels with 2-5 digits and optionally followed by a letter. We recommend a minimum of 3 digits to prevent confusion with other numbers in your specification. Our reference label results are presented as a table with three columns:

  • Reference labels — The first column lists the reference labels we found in your application and/or figures. Some of these may not actually be reference labels. You can double click a reference label to immediately see the paragraphs from the spec where you used the reference label.
  • Specification — The second column lists the text we found in your specification for the reference label along with a count of the number of times each phrase appeared. If the cell is red, we didn't find the reference label in the spec or it was used for different things. If the cell is yellow, the reference label may have been used inconsistently. You can double click the reference label text to immediately see the paragraphs from the spec where you used the reference label with the text.
  • Figures — The third column indicates whether the reference label appeared in the figures and, if it does, the figure numbers on the same page as the reference label to help you quickly find the reference number in your figures.

We also check to make sure that you have used your figure numbers consistently. We look for all possible ways you may number your figures (e.g., "Fig. 1", "Figs. 2A-C", "Figure 3", etc.). Our figure number results are also presented as a table with three columns:

  • Brief Desc. — The first column lists the figure numbers found in the "Brief Description of the Figures" section of your application.
  • Spec. — The second column lists the figure numbers we found in your specification after the Brief Description of the Figures.
  • Figures — The third column lists the figure numbers found in the drawings.

Each figure number should appear in all three columns, and the red cells indicate when a figure number is missing in a location.

Claim Sets

We can also proofread just a set of claims. For example, you may draft a set of claims and want to proofread them before writing the specification. To do this, proofread the claim set the same way you would proofread a patent application. To identify the claims, we just need a paragraph before them with only the word "Claims" or a phrase like "What is claimed is:" (any capitalization).

Office Action Responses

We can process your office action response with or without your patent specification.

Expected Format

The content of office action responses is more varied so we look for the start of your claims and the end of your claims.

  • Start of Claims — We search for the first paragraph of your office action response that looks like a claim. To avoid mistakes, we only look for a claim with a claim status indicator (e.g., "Currently Amended") so you'll get an error if your claims don't have status indicators.
  • End of Claims — For the end of your claims, we look for typical section headings (e.g., "Remarks" or "Amendment to the Specification") or a paragraph after a page break that doesn't look like a claim.

Analyze Claims

The processing here is mostly the same as for patent applications. Here are the main differences:

  • Specification — To check the word and phrase support of your claims, we need your specification. We thus allow you to upload the MS Word doc of your application. Though, if you don't provide the application and your application is public, we'll go get it for you! So if your application is public, just leave the box for uploading the application empty. If you don't provide your application and it isn't public, then we can't check word/phrase support but we'll still look for numbering and antecedent basis errors.
  • Claim Status Indicators — We'll let you know if you've amended a claim but forgot to change the status to currently amended or vice versa. We treat a claim with overstrikes, underlines, double brackets, or track changes as being amended.
  • Claim Summary — Many people start the remarks section with a summary of claim changes (e.g., "Claims 1-21 are pending. Claims 3, 8, and 13 are amended. Claim 21 is new."). We generate this text for you so you can copy and paste it into your Word doc.

Microsoft Word Add-In

Our proofreading tools are also available via a Microsoft Word add-in for Word 2016 or later. Our add-in works on both Windows and Mac and also for both purchased and subscription versions of Word (purchased Word requires Word version 16.0.4266.1001 or later).

We have a version of our add-in for processing on Google Cloud (Patent Bots GCP) and a version for processing on Amazon Web Services (Patent Bots AWS).

There are three ways to install our add-in:
  1. Your IT department can install for a group of users or for your whole firm using the Office admin tools. See instructions here and here.
  2. You can install for yourself from the Microsoft store (Patent Bots GCP and Patent Bots AWS).
  3. You can install for yourself within Word: Insert tab in the ribbon (not the menu) -> Add-ins button -> Get Add-ins-> search for Patent Bots.

Our add-in provides easy access to the patent proofreading tools described above. When you start proofreading, we will open a tab in your default browser with the proofreading results. The proofreading results are otherwise identical.

Our add-in requires that your patent document be smaller than 4 megabytes. Very few patent documents are larger than this, but if you have a larger one, then you need to start proofreading from our website.

Our add-in also provides access to our Ref Label Tracker decsribed below.

Print Proofreading Reports

You can use the print functionality of your browser to print our proofreading results to paper or to a PDF file. When printing, you need to select the print option for "Background graphics" in order for the highlighting to appear. In the Chrome print dialog, you can find this option under "More settings" and it looks like this:


If your patent application or office action response doesn't match our expected format, then you may get a processing error. If this happens, please feel free to contact us for assistance.

We can usually process patent documents in about 2 seconds. In some cases, our processing may take longer or reach our maximum processing time of one minute and return an error. Here are a few instances in which this may happen with suggested solutions.

Very Long Specifications or Claim Sets

We have successfully processed patent applications with several hundreds of pages and several hundred claims. If your processing is taking too long or if you receive errors, please try using the "Analyze Claims" button instead of the "Analyze All" button.

Certain PDF Files

Some PDF files are very slow to process. This is more likely to happen with a PDF received from a draftsperson than from PowerPoint or Visio. If your processing is taking too long or if you receive errors, please try processing without your figures.

Too Many Multiple Dependent Claims

Patent Bots works with multiple dependent claims, but be careful to not have too many. If you have claims like this:

1. A method for foo.
2. The method of claim 1 wherein ...
3. The method of any of claims 1-2 wherein ...
4. The method of any of claims 1-3 wherein ...
20. The method of any of claims 1-20 wherein ...

then the total number of claims is about 500,000, and this exceeds our current processing capabilities!

Drafting Assistant

In addition to proofreading, we are working on tools to make you more efficient in drafting patent applications and office action responses. Our first drafting tool is our Ref Label Tracker, and we are working on additional drafting tools.

Ref Label Tracker

Our Ref Label Tracker identifies reference labels and their corresponding text in real time while you are drafting your patent application. We show a list of ref labels and corresponding text in a side panel for easy reference. This makes it easy to identify a needed ref label so you don't have to manually search in your Word doc. The list of ref labels updates automatically every 2 minutes (to avoid overloading our servers).

Our Ref Label Tracker is available within our Word add-in. See our instructions above for how to install our Word add-in.


We have a couple of utilities to handle common, routine tasks.

Convert Track Changes

This is a simply utility for people who use MS Word track changes in editing their claims. We process your Word doc and convert the track changes to underlines and overstrikes. Note that we convert ALL track changes to underlines and overstrikes so you should resolve track changes outside of the claims before using this utility.

Generate Office Action Response Shells

Before you start working on an office action response you need to update the claims from the application or a previous office action response. We automate this process for you.

  • Create from Application — For your very first office action response, you need to take the claims from the patent application and add a status of "Original". We extract the claims from your patent application and add the Original status to each claim.
  • Create from Previous OA Response — For subsequent office action responses, you need to modify the claims from the previous office action response. We extract the claims from your previous office action response, remove underlines, remove text with strikethrough or double brackets, and change a claim status of "Currently Amended" or "New" to "Previously Presented".

We are working on expanding our shell generation. In the future, we will allow you to use your own templates and also extract relevant information from the USTPO (e.g., art unit, examiner name, etc.).


Our PatentPlex is a one-stop shop for information about a particular patent. Just enter a patent number to get started. Our PatentPlex includes patent family trees, prosecution timelines, and easier access to PAIR.

Patent Examiner Statistics

Our patent examiner statistics are mostly self explanatory. The statistics relating to appeals are available with a subscription and the other statistics are available for free to all logged in users.

We are working on adding more statistics. Please let us know what other statistics you would like to see, and we'll do our best to include them.

We update the statistics about every two months, and our data set is all public applications filed in 2010 or later.