What AI should I recommend to a non-technical friend? The script

What AI should I recommend to a non-technical friend?

Recommend one app that picks the model for them. One login, one text box, no choices, about $5 a month. That is the whole script — and the reason it is the whole script is the story of why Plainly exists at all. This is the article I wish someone had handed me before I spent years sending friends links that didn't work for them.

The moment that started this

A while back I wanted to hand a non-technical friend access to DeepSeek-class models. The models are genuinely great — strong enough for everything an everyday person does, at a cost that makes a $5 subscription work. And my friends, like most people, do not want to spend real money on AI.

There was just one problem: there was nothing clean to send them. The open-source labs target technically adept people. Their apps assume you know what a model is and enjoy choosing between them. There was no polished website where a normal person could just use these models. I could not point my friends at a good product using them, because it did not exist.

That gap is the reason I built Plainly — great models, cheap cost, clean interface. This is not a hypothetical recommendation framework I am dressed up for SEO. It is the founding motivation of the product, and the reason "recommend it to a friend" is built into how it grows: every subscriber can hand a friend a free month, because the moment of recommending is the moment this category lives or dies.

Why the wrong recommendations fail

Most recommendations to non-technical people fail in predictable ways, and knowing the failure modes tells you what a good one looks like:

• "Just use ChatGPT" — lands them on a free tier with caps and a reduced model, or a $20 price they baulk at. They use it twice, hit a limit mid-task, and conclude AI isn't for them.
• "Try DeepSeek, it's free" — the models are excellent, but the app targets technical users. A non-technical friend opens it, sees jargon and rough edges, and closes it. The gift bounces.
• "You should try a few and see what you like" — the single worst option. Comparison is precisely what they asked you to save them from. Every model comparison is homework, and homework is why they asked you instead of searching themselves.

The pattern: every failed recommendation hands the technical burden back to the person least equipped to carry it. A good recommendation absorbs it completely.

The three questions your friend will ask

Whatever you recommend, it needs clean answers to these — this is the checklist I run before sending anyone a link:

1. "What do I type?" — Anything, in normal words. If the tool needs prompt technique or special phrasing, it is the wrong tool. The test: would their actual first message, typed exactly as they'd say it out loud, get a useful answer?
2. "What does it cost?" — One number. $5 a month. No credits, no tokens, no tiers, no "free with limits" asterisk. Complexity in pricing is the fastest way to lose a non-technical user, because every option is another decision they have to research.
3. "What if I press the wrong thing?" — They cannot. There should be nothing to press wrong. No model picker, no settings that matter, no way to break it. If the app has a settings screen, your friend should never need to open it.

Run any recommendation through those three. If it clears them, send it. If it doesn't, keep looking — the right tool exists, and sending the wrong one costs you the trust the next recommendation would have needed.

The script, in my actual words

When someone asks me what AI to use, I say something like:

"Use Plainly. It's $5 a month. You type a question like you'd text a friend, and it answers. There's nothing to set up or choose — I've already got the technical part handled. Here's the link."

Notice what the script does not contain: model names, comparisons, benchmarks, setup steps, explanations of how anything works. A recommendation that requires a briefing document is not a recommendation. The reference point that lands with everyone is the phone: if they can text a friend, they can use it.

The script works because it does three things at once: it gives them one action (use this), one number ($5), and one reassurance (I have handled the technical part). Those are the three things a non-technical person actually wants to know. Everything else is noise you are politely declining to add.

Why the recommendation itself is the product

A technical person vouching for a tool is the trust layer non-technical users are missing. They have heard of ChatGPT, they suspect it costs $20 and involves knowing things, and so they use nothing. One clean recommendation from someone they trust clears all of it — the price anxiety, the setup fear, the "am I doing it right" doubt. This is why the technical person is the evangelist and the non-technical person is the customer: the recommendation is doing the work marketing can't, because it comes from someone with no reason to oversell.

If you are the technical friend doing the telling, how Plainly handles models explains what you are vouching for, and is a $5 AI good enough covers the price question before they ask it.

The honest limit

If your friend turns out to be a power user — wants API access, wants to choose models, follows the labs' release cycles — do not force the simple tool on them. They will feel condescended to, and they would be right. But those people are rare. In my experience everyone else is quietly waiting for someone to just tell them what to use, and deeply relieved when someone finally does.