Battle of a Billion AI Chatbots

Happy Monday Non-Techies,

ChatGPT, Claude, Grok, Deepseek, Copilot, Gemini, LeChat…there seems to be a billion AI chatbot options right now. Even as someone whose job it is to keep up, lately it’s been really confusing with so many new things coming out, seemingly every day.

The thing is, these chatbots don’t all work in the same way - they all have different features and functions, so I’m going to try and lay it all out clearly for you today.

Before we start, a couple of quick AI for Non-Techies updates:

• Our popular Become an AI Trainer bootcamp is coming up next week, running 4-6 March. And it’s only £72/$97 for 3 cram-packed days, 90 mins a day, live, with me. Grab your ticket here.

• Join the team at AI for Non-Techies…there’s just a few days to apply for our Academy Community & Growth Manager job - get your form in before Friday to be considered.

Now, let’s get this AI chatbot thing cleared right up.

There are 4 Types of AI Chatbot

Let’s clear up the LLM thing first. An LLM is a Large Language Model, and we can think of it as a sort of engine, it’s the power behind the AI chatbots and the reason they work. So ChatGPT isn’t an LLM technically, it’s the chatbot that uses the GPT4 engine (that’s the LLM bit) - the chatbot is the thing we type into to access it. Some chatbots share the same engine - Copilot and ChatGPT, for instance, both use the GPT4 engine.

So we have LLMs as the engine behind the chatbots.
And then we have the chatbots themselves.

Right now, we have 4 different types of AI chatbot. I just pulled this graphic together on Canva for you to try and break it down:

Type 1: General Chatbots

These are your everyday use chatbots, and can happily take care of most use cases: content creation, summarisation, simulating conversations, etc. You can also analyse simple data conversationally with these - ChatGPT and Claude have great data analysis modes.

These general chatbots include being able to create your own custom versions like building Custom GPTs, Claude Projects, etc. I’ve got probably 20 of these right now to help me create bespoke proposals, design webinars, write blogs, improve my anxiety and lots more.

You can achieve a LOT with these alone - I’ve tripled my business size with general chatbots. These are the ones that most people need to focus on - though it’s easy to be distracted by the shiny new Research and Reasoning chatbots.

Type 2: Research Chatbots

These are new, and have come out in the last month or so. Although we’ve been able to do research using ChatGPT and Perplexity, we haven’t been able to go very deeply into it. Now we can - they claim it’s like having a PhD level researcher available, 24/7.

It’s designed for:

• Market research

• Product research

• Academic research

The results you get are comprehensive, fully cited research papers, comparable to something created by a professional researcher. The level of depth is actually pretty dazzling.

They take a lot longer than general chatbots to “think”, but usually that means a much stronger answer. It’s constantly adapting its plan as it gets more and more into it. Some of the research models take 10-15 minutes to give you a response back - this is not a quick tool - but the results are thorough.

Here’s OpenAI’s video explaining Deep Research:

They are directly linked to the below type of chatbot, in that they’re a type of Reasoning chatbot designed specifically to research. So…what the heck is that then?

Type 3: Reasoning Chatbots

These are also new, though they’ve been around a little bit longer than the research chatbots above. They’re kind of like one of those very clever people that struggles with common sense stuff - the more complex the query, the better it seems to do, but try it with something basic and it’s poor.

It started with the confusingly named ChatGPT o1 (3.5, 4, 4o, er…o1, anyone?), though it turns out the naming convention was because it’s a totally different type of chatbot. They were starting from scratch with something different: the ability to reason - hence the o1.

The most well-known reasoning chatbot is Deepseek R1. Deepseek has been around since Jan 2024 with their general chatbot, but it was their reasoning version - R1 (R for reasoning) that caught everybody’s eye - well, that and their insane marketing strategy.

It caused a lot of confusion, as certain major AI influencers (no naming names) said people should ‘delete ChatGPT, Deepseek is here’. A few problems I had with that:

1. Deepseek has been around for a year, you’re referring to Deepseek R1

2. ChatGPT is a general chatbot, Deepseek R1 is a reasoning one, they do different things, you need both

So what do you actually do with them?

Use them for “normal” use cases and the results are often disappointing. They’re designed for more complex queries only - specifically science, technology, engineering and maths (STEM) topics, though I’ve found them very useful for detailed training programme planning and pricing strategies. I also used it to help me create a content marketing strategy and the results were incredible.

Type 4: Aggregate Chatbots

The fourth type of chatbot right now is the aggregate site.

These sites allow you to use multiple different models at once, usually for the price of only one of them, about $20/month.

Pros

• Price - save a ton of money on different subscriptions

• Compare the results of different models for the same prompt

• See what suits you best before you buy

• They are very quick to get the latest chatbot models

Cons 

• Functionality - you don’t get access to all features of every model

• It can be a bit confusing jumping between different models

• They don’t make it that easy to compare results

• Often, it’s hard to know which model you actually want!

One of the most popular examples is you.com. You get this long list of different models - every single one I can think of - and you pick which one you want to use.

Another great aggregate site is Launch Lemonade, which offers a lot more than the different models - it calls itself ‘Canva for AI’. They actually explain what each model does, which is helpful:

Meet the Founder of Launch Lemonade and see it in action in our new members-only session: Cool Tools with Nick session in our Academy, starting on Wednesday this week.

OK, So Which is Which?

Bear in mind this changes quickly, so this is only relevant for where we are right now (23 Feb 2025 when writing this!):

General Chatbots:

ChatGPT - the popular, innovative one
Grok - high memory, all-rounder, no censorship
Claude - the creative, human-sounding one
LeChat - the underdog that just got a major upgrade
Gemini - Google’s great all-rounder
Copilot - great integrations, disappointing functionality

Research Chatbots:

Grok - click ‘research’ in Grok 3, now out
Gemini - now has Google’s Deep Research
ChatGPT - Deep Research, available on its Pro account*

*You can also this via Perplexity Pro

Reasoning Chatbots

Deepseek - their R1 model - available directly** or via Perplexity
ChatGPT - O3-mini (replaces their o1 model…what happened to o2?)
Grok - click ‘think’ in Grok 3

I’m sure this will change pretty soon but that’s where we are right now. I think we’ll see both research and reasoning functionalities within all of the general chatbots very soon, a bit like Grok 3 has done.

Hopefully now you’re feeling a bit clearer about all these new chatbots and new functions.

That’s it for me this week. I’ll be filming our new Build an AI Agent learning track and Ask a Tech Lawyer session for our Academy this week - all great learning for me, too.

Oh, and I’ve decided I’m doing a complete refresh of the newsletter structure featuring news, opinions, prompts and much more. Watch out for that in your inboxes very soon.

Speak very soon,

Heather

PS Learn more about AI and meet live with me every week in my affordable AI Academy at just £25/$31 per month.