Authenticity is Everything
Has anyone tried this Verimarx Ai tool that's advertised on this site?
Was wondering if anyone here has used it here and had it work well? I can't find much information on this tool, supposedly it detects forgeries, just wanted thoughts on it.
Replies to This Discussion
-
Permalink Reply by Dominick Lisi on
-
I appreciate the thoughtful feedback here and wanted to add a little more context as one of the developers.
At this stage, VeriMarx is very much a research and collaborative project, not an authentication service. As the FAQ states, it does not authenticate, certify, or guarantee an autograph. The purpose is to provide analytical insight that may help collectors decide whether an item is worth pursuing further authentication.
The focus of the analysis is not simply “real vs fake.” The goal is to explain the why — identifying potential forgery indicators, common forgery techniques, and inconsistencies, and then comparing those observations against known authentic representations where applicable. The resulting report is meant to be educational, highlighting characteristics collectors may want to examine more closely.
Equally important, the system is still very early in development (about 4–5 months old) and the current phase is intentionally open so that collectors can experiment with it, critique it, and help improve it. Feedback from experienced collectors — especially communities like this one — is exactly what helps refine the analysis.
Regarding the concern that tools like this could help forgers: the reality is that forgery techniques have been documented in books, articles, and forums for decades. Our goal is to level the playing field for collectors, many of whom don’t have access to large exemplar libraries or years of experience studying handwriting characteristics.
Ultimately, VeriMarx isn’t meant to replace collectors’ knowledge, community opinions, or third-party authenticators. If anything, the goal is to encourage more informed decision-making and help collectors better understand the kinds of indicators experts already look for.
Constructive criticism is genuinely helpful at this stage, so if anyone here tries it and sees areas where it can improve, I’d absolutely welcome that feedback.
-
-
Thanks for the background information.
In regard to your comment that it compares by using "known authentic representations where applicable", where are those coming from?
-
-
Quite frankly....from you. As well as sources that are open and available. Also, the comparative analysis has proven far less valuable and accurate than the forgery indicators.
-
Permalink Reply by Eric Keith Longo on
-
It would be interesting and useful to see some examples of your Jackie Gleason exemplars through the years 1945 to 1985.
-
-
It might be more informative for you to try one that you have or have access to. I think you would find the analysis interesting.
-
-
I have 57 authentic exemplars posted here from 1936 to the 80's. I was curious to see what you are using for exemplars.
-
-
I would recommend running 3 to 5 examples from different errors through the application to see what the analysis suggests. It should tell you if it’s an adequate exemplars to compare against
-
-
One way street? I want to see what you are using. I believe I know what I am doing with Gleason in each era. You can't post any of your exemplars? But you want me to submit mine?
-
Permalink Reply by Steve Cyrkin, Admin on
-
Why don't you submit 5-10 a day? That'll get them into the database it learns and recognizes trends from.
-
-
If the thing is reading this site my exemplars are already posted. I gave a link. I am curious to see what Verimarx is using, particularly in the final secretarial phase which represents the majority of what is out there certified as authentic. PSA still gets them wrong - and right too. The Honeymooner's Facts Page still contains more secretarials and forgeries than authentic examples.
-
-
So when someone submits an autograph for analysis that goes into the database as an exemplar?
Yikes.
-
-
In my original post on this thread I expressed some skepticism that the AI might just be pulling exemplars of all kinds (real and fake) from the internet to use for comparison, and you insisted that it does not work that way.
But when I then asked where you get the exemplars for comparison you replied: "Quite frankly....from you. As well as sources that are open and available." Isn't that sort of the same thing that I mentioned previously?
It seems that you are not using a database of established authentic examples. So, are you scouring the internet for examples to use for comparison? How do you separate the good from the bad when you harvest the examples?
These are just honest questions about the process, since I still don't feel confident about the source of the exemplars.
Sign up for our Newsletter!
Photos
© 2026 Created by Steve Cyrkin, Admin.
Powered by
Badges | Report an Issue | Privacy Policy | Terms of Service
