Charge $500 for 600,000 words? Hear me out. – 25 February 2026 (AI Mastery for Translators Insights)

NAVIGATING THE PARADIGM SHIFT

Charge $500 for 600,000 words? Hear me out.

There’s something psychologically powerful about offering an “unlimited” amount of a service for a fixed fee. This removes a major friction point from the transaction; the buyer doesn’t have to balance priorities. They can just take what they want.

My mobile phone contract offers me unlimited data use. But this isn’t actually true. Check the fine print, and you’ll see that “unlimited” is subject to a “reasonable use” clause. There actually is a limit. But they’ve defined that limit at a point so high I would only reach it if I were being “unreasonable”, like leaving unwatched YouTube videos streaming 24/7, or automating the download of huge numbers of files from online document repositories.

But since I’m not doing any of that, for all intents and purposes, my data plan is unlimited for me. And because not having to self-ration is a valuable benefit in and of itself, I’m happy to pay more for an unlimited plan than I would for a data plan that actually matches the amount of data I’m really using.

What if you could offer your clients “unlimited” translation?

Last week I suggested a cheap raw machine translation service that you might offer to your direct clients at 5-10% of your full rate. Even though ChatGPT translation is now free already, I justified this charge based on the better job you can deliver. I listed a number of ways your raw machine translation, done with the GAIT workflow in your CAT tool using CotranslatorAI, can be better:

So let’s back up and rethink the “retainer” approach I also recommended last week. What if you package in “unlimited” raw MT along with the full-human and turn-key MTPE services you’re already offering at per-word rates? Now your client can get everything translated.

Everything? That’s literally all the information in the world.

No, it doesn’t have to be. Include a “reasonable use” clause and define it as the most content one person can reasonably read in a month… That’s your “unlimited”.

Suppose the average person can read 10,000 words/hour. Surely two hours a day is the most anyone could possibly sit around reading your translated stuff day in and day out. That’s sixty hours per month, or 600,000 words… Even if your client actually does request that much work, you can probably process it with the GAIT workflow in 20 hours or so. This is a worst-case scenario.

What would you need to earn to justify 20 hours of work a month? $1000? Charge the client $500. Why? Because they probably won’t actually need that much. And your full retainer is actually $2,000/month, including the $1,500 your client has to commit to for per-word work before they’re even eligible for the “unlimited” raw machine translation.

Besides regular income, what else do you get in return? Your client’s mindset about you changes. They literally send you everything. You’re the go-to person now. This is value you’ve never offered before, and it’s the starting point for becoming part of the team, not just another “vendor”.

Sound risky? Fine. Your client’s going to feel it’s a risk too. So start with a three-month contract and renegotiate after that.

Am I sure this is the way you should run your business now? No, of course not. We’re feeling our way through this paradigm shift together. But I don’t think this idea is bonkers either.

What do you think? Can we share ideas and come up with new and better approaches?

EFFECTIVE PROMPTING

If your translation process is “too easy”, you’re doing it wrong.

After last week’s newsletter, one of you reached out to push back on my idea about selling raw MT. He took issue with my suggestion that a translator should aim for 30,000 words an hour for raw machine translation.

The question was simple: why not just set up the anchor prompt once and automate the rest with programming? Wouldn’t that mean you could translate gobs of content overnight, segment by segment, with no effort at all?

It almost seems too easy… and it is.

The key point he missed here are the unique advantages you bring to the process by staying engaged.

The biggest one is context leverage. Even if your anchor prompt isn’t perfect, you can already achieve noticeably better translation quality and internal consistency just by handling multiple segments at a time. Keeping translations within the same context window helps you produce a more coherent final product. This is a huge quality boost, and it simply can’t be automated. Especially if you’re using your CAT tool (which you should be for many reasons) you’ll need to babysit the AI or risk real chaos.

There’s another challenge as well: as you make the most of context, the context window grows, and eventually the AI will start losing reliability. When that happens, you’ll need to transition to a new context window, and this is the perfect time to update your anchor prompt to reflect what you’ve learned in the first cycle. Layering this in means each new content block has the highest shot at both consistency and quality.

Throughout the process, there are other touchpoints where your engagement adds value. Sometimes you’ll notice the anchor prompt no longer matches the kind of content that’s coming through, or you’ll spot something odd in the output that needs intervention. Even if you aren’t checking every word, your presence and strategic tweaks keep the process aligned and humming.

And here’s a practical concern: if you’re just running the workflow fully automated, segment by segment, you lose out on consistency, context, and, ironically, you’ll likely spend more on tokens if your prompt is robust. By following manual GAIT workflow principles, you can actually use the top models while still keeping your costs down.

Finally, remember that automating the whole thing with your own programming is not a unique value proposition… this is precisely what the Localization Illuminati are doing, and they’ll always have more resources than you to keep this cycle going. You can’t win by playing their game.

Maybe you think the difference in cost or quality is small, maybe not. But your only real value as a human translator is the value you add… the things that automation can’t deliver. That’s the only way forward if you want to keep doing what you love on your terms.

Keep your eye on human value, or it might be time to find something else to do. To me, those are the only real options on the table.

COTRANSLATORAI TIP

Delete your API key backup right now

When you use CotranslatorAI, you need an OpenAI API key. You might be wondering how safe your key is and whether you should back it up somewhere, just in case?

The answer is straightforward. CotranslatorAI stores your API key in encrypted form on your computer. That means even if someone tries to read files off your hard drive, your key won’t be visible or usable. When CotranslatorAI connects to the AI server, it sends your key over a secure HTTPS connection, which is standard practice for protecting sensitive information online.

But since OpenAI doesn’t let you view your API key a second time (after you first generate it), maybe you should keep a backup copy in a text file for emergencies?

Actually, there’s no need for this. If you ever lose your key or think it might be at risk, you can just generate a new one instantly. Saving copies of your API key outside of CotranslatorAI only adds risk, not safety.

So once your key is in CotranslatorAI, you can delete your backup and forget about it. This is a simple step to keep your API key secure and your AI work safe.

We’re all still figuring out this new landscape together. What’s working well for you lately? I’d love to hear about it.

Steven (for the CotranslatorAI team)

Steven S. Bammel, PhD

Korean-to-English Translator
Co-Creator of CotranslatorAI

PS: Check out the What’s New page for the lastest onboarding resources.