Google has spent an enormous amount of engineering effort making Gemini one of the most capable AI systems available to consumers. Google has also spent that same period making it nearly impossible for a normal person to explain what Gemini actually is.
Both things are true at once, and the second one is quietly eating the first. That’s the argument TechCrunch made this week, picked up by Yahoo Tech and Bitcoin World, and it’s the rare piece of AI commentary I didn’t want to argue with.
A naming problem is a product problem
I review these tools for a living. I keep notes. And I still occasionally have to stop and work out which product someone is talking about when they say “Gemini.” Is it the model? The app? The thing bolted into Gmail? The thing that replaced Assistant on your phone? The version number attached to whichever of those you happened to open?
That confusion isn’t cosmetic. When a user can’t name what they’re using, they can’t recommend it, can’t search for help with it, and can’t tell whether the thing that impressed them last month is the same thing disappointing them today. Word of mouth dies without a noun.
Compare that to the competition, where the product name and the company name have effectively merged in the public mind. You don’t need to explain the relationship between the app, the model, and the brand. There’s one door. Google has built a hallway.
This is not just a Google problem
The TechCrunch framing is right to widen the lens, because the whole industry has developed a taste for names that mean nothing to anyone outside the lab. Version numbers that skip. Suffixes that imply speed, size, or reasoning depth with no consistent logic. Model families that split into variants faster than anyone can document them.
Inside an AI company, this makes a certain sense. Engineers need to distinguish between checkpoints. Product managers need to signal capability tiers. But the naming conventions of a research org leaked into consumer marketing, and nobody stopped it.
The result is a category where the most technically advanced products on earth are marketed with the clarity of a printer model number. My working theory is that everyone shipped so fast that branding became the last item on the list, and by the time anyone looked up, the confusing names had already shipped to hundreds of millions of people.
What confusion actually costs
Here’s what I see in practice when I test these tools and talk to people using them:
- Users stay on older, weaker versions because they don’t realize a newer one exists or applies to them.
- People blame the wrong product for a bad experience, which poisons the brand they can actually name.
- Comparison becomes impossible for anyone who isn’t already an enthusiast, so purchase decisions default to whatever’s familiar.
- Support and documentation fragment, because there’s no single term to search.
None of that shows up in a benchmark score. All of it shows up in adoption.
The pattern shows up elsewhere too
Google’s week gave us two other stories that rhyme with this one. The company rolled out a new way for publishers to fight AI-driven traffic losses, which is Google trying to solve a problem Google’s own products created. And TechCrunch’s Pixel 11 Pro XL review landed on a familiar verdict: faster cameras, iterative upgrade.
Different products, same underlying issue. When the improvements are real but incremental, the story you tell about them matters more, not less. An iterative phone needs sharp positioning to justify itself. A model family that gets meaningfully better every few months needs a name people can hold onto, or the improvements go unnoticed.
Google has historically treated naming as a thing that can be redone later. Sometimes that works. Chat apps got renamed into oblivion and nobody outside tech noticed, because nobody outside tech was using them. Gemini is different. It’s in front of an enormous number of people right now, and every confused user is a small permanent loss.
My take
The honest read is that Google is winning on capability and losing on legibility, and legibility is the part that compounds. You can ship a better model next quarter. You can’t easily un-confuse a billion people about what your product is called.
If I were advising them, I’d pick one consumer-facing name, retire the variant soup from anything a normal user touches, and keep the technical designations where they belong, in the developer docs. That’s not a marketing flourish. It’s the difference between a product people talk about and a product people vaguely remember trying.
The rest of the industry should take notes. Right now, the naming is the weakest part of nearly every AI product I test, and it’s also the cheapest thing to fix.
🕒 Published: