How to Use Artificial Intelligence to Validate Translations

Artificial intelligence (AI) can be a powerful ally for translators — as long as it is used with discernment. More than simply “translating,” it can serve as a valuable tool for terminological and conceptual validation.

In a previous post, I explained that it is important to use ChatGPT with caution. In this article, I present a real example involving a common legal term in Brazilian documents: “reconhecimento de firma”.

The problem: when a literal translation is not enough

While translating a procuração (Power of Attorney), I came across a section titled “RECONHECIMENTO DE FIRMA,” followed by a statement in which the notary declares that:

  • the signatory appeared in person,
  • presented identification, and
  • confirmed that the signature is their own.

My initial solution was to translate it as:

SIGNATURE NOTARIZATION

But something felt off.

A plausible — but problematic — suggestion

While discussing the term in a translators’ group, someone suggested:

NOTARY CERTIFICATE

At first glance, the expression seems appropriate. However, it raised an important question:

Doesn’t “certificate” refer to a separate document that would be attached to the power of attorney?

With that in mind, I considered changing the suggested translation to notary certification, but for some reason, my intuition told me there was a mismatch between reconhecimento and certification. Reconhecimento is an acknowledgment, so from that point on, I began researching online.

The research and the proposed translation

While searching for references, I found multiple occurrences of the term acknowledgment in the notarial context, as well as legal definitions such as this one:

“Acknowledgment” means a declaration by a person that the person has executed an instrument for the purposes stated therein and, if the instrument is executed in a representative capacity, that the person signed the instrument with proper authority and executed it as the act of the person or entity represented and identified therein. (Wisconsin Legislature website).

Based on this, I arrived at the following proposal:

NOTARIAL ACKNOWLEDGMENT

A quick Google search for “notarial acknowledgment” returned approximately 29,000 results and 162,000 for the alternative spelling “acknowledgement,” which attests to its popularity, though not necessarily its technical accuracy.

Where AI came into the process

At this point, I turned to AI not to “discover” the translation, but to critically validate the solution I had found.

I provided:

  • the original term (reconhecimento de firma),
  • a description of the procedure (appearance, identification, signature confirmation),
  • a citation from my reference source, and
  • the translation I had proposed (Notarial Acknowledgment).

The response confirmed that:

  • the elements described correspond to the concept of acknowledgment in U.S. notarial law;
  • the proposed translation is technically appropriate;
  • alternatives such as “notarization” or “certificate” are too generic or imprecise in this context.

What AI actually did

In this case, AI did not provide the initial solution. Its role was to:

✔ validate the functional equivalence between legal systems
✔ confirm the terminological adequacy of the proposed solution
✔ reinforce why other alternatives were less precise

In other words, it acted as a qualified second opinion.

A detail that makes all the difference

The analysis also highlighted an important nuance:

in Brazil, reconhecimento de firma can be performed por autenticidade or por semelhança.

In my case, it involved personal appearance — that is, por autenticidade. This justifies the use of acknowledgment.

Click here to read my full interaction with the AI.

Best practices when using AI to validate translations

  1. Do your own research first
    AI works best as a validator, not as a starting point.
  2. Present your hypothesis
    Instead of asking “how do I translate this?”, ask “is this translation appropriate?”.
  3. Provide detailed context
    Especially in legal texts, the procedure matters more than the isolated term.
  4. Use AI as a second opinion, not as the final authority
    Responsibility still lies with the translator.
  5. Question answers — including those from AI
    Critical validation is an essential part of the process.

Conclusion

Artificial intelligence does not replace the translator — but it can significantly elevate the quality of their work when used properly.

In this example, the solution did not come from AI, but it was confirmed and strengthened by it.

And that illustrates its best use: not as a source of ready-made answers, but as a tool for intelligent validation.

PS: Additional research confirms the use of “notarial acknowledgment” in accordance with the notarial laws of Arizona, California, New York, and Wisconsin, the latter already mentioned.