AI search & GEO
Types of AI: What Each One Does and Which to Use in Marketing
The 3 and 4 types every course lists, boiled down to two tables, and the map nobody gives you: which type of AI handles each marketing task, what number to track and which tools to try.
The types of AI are grouped three ways: by capability, by how they work and by the task they do. The first two show up in every course, but for your marketing only the third one matters: generative, predictive, conversational, computer vision, recommendation and agents.
In the US, "types of ai" gets about 18,100 Google searches a month, and people ask AI chatbots the same thing roughly 8,500 times a month, according to DataForSEO. Most of the top 10 repeats the academic version: narrow AI, general AI, superintelligence. Good to know. Useless for deciding which tool to open on Monday.
So here are the two textbook classifications in two tables, followed by the map you'll actually use: which type of AI handles each task, what number to track and which tools to test. It's the foundation for everything we cover in AI search.
On this page
What are the different types of AI (and what isn't a type)
There are three ways to sort the different types of AI: how capable they are, how they work and which job they do for you.
The types of AI are the categories used to group AI systems by what they can do. They're classified by capability (compared with a human), by functionality (whether they remember and how much they understand) and by task (create, predict, converse, see, recommend or act). One tool can fit several categories at once.
What isn't a type: machine learning, deep learning and natural language processing are techniques used to build almost every system, which is why the machine learning vs AI debate is really about method versus goal. Model families like LLMs and diffusion models, which is usually what people mean by types of AI models, are the engines underneath. And ChatGPT is a product, not a type. Mixing techniques, products and types is a mistake you'll find even in the top 10 on Google.
| Classification | Types | Does it help you pick a tool? |
|---|---|---|
| By capability | Narrow, general, superintelligence | No: only narrow exists |
| By functionality | Reactive, limited memory, theory of mind, self-aware | No: nearly everything is limited memory |
| By task | Generative, predictive, conversational, computer vision, recommendation, agents | Yes: this is the one you decide with |
The 3 types of AI by capability: narrow, general and super
By capability there are 3 types of AI, and only one of them exists: narrow AI, which does one specific job very well.
Narrow AI (also called weak AI or ANI) handles one task or a small set of them: translating, recommending, writing text. Artificial general intelligence (AGI, or strong AI) would match a person at any intellectual task. Artificial superintelligence (ASI) would beat us at all of them. The last two are hypotheses, not products.
If someone pitches you "an AI that does everything," they're selling several narrow AIs bundled together.
The 4 types of AI by functionality
By functionality there are 4 types of AI, ranging from machines with no memory at all to machines that would be self-aware.
This framework was popularized by Arend Hintze, then a professor at Michigan State University, in a November 2016 piece for The Conversation.
| Type | How it works | Example |
|---|---|---|
| Reactive machines | React to the present, no memory | Deep Blue, which beat Kasparov at chess in 1997 |
| Limited memory | Learn from data and use recent context to decide | ChatGPT, spam filters, cars with automatic braking |
| Theory of mind | Would understand emotions and intentions | Doesn't exist |
| Self-aware | Would be conscious of themselves | Doesn't exist |
What are the 7 types of AI?
The 7 types of AI are simply the two lists combined: the 3 by capability plus the 4 by functionality. That's why you'll see guides promising "7 types" and others "4 types": same content, different packaging.
What type of AI is ChatGPT?
ChatGPT is narrow AI with limited memory, and by task it's both generative and conversational. It proves the types aren't sealed boxes: the same tool writes an ad and answers a customer. That's why choosing by task beats choosing by label.
Types of machine learning: supervised, unsupervised and reinforcement
Nearly all the AI you use learns through machine learning, and machine learning comes in three types, depending on how the model learns.
| Type | How it learns | What it solves in marketing |
|---|---|---|
| Supervised | From examples that already include the answer | Predicting who buys, who churns or how much they'll spend |
| Unsupervised | From unlabeled data, looking for groups | Segmenting customers by behavior |
| Reinforcement | By trial and error, with rewards and penalties | Adjusting on the fly, like sending more traffic to the winning ad |
Here's a made-up example for the guide: an online running-shoe store with 5,000 customers. Flag who bought a second time and who didn't, and a supervised model learns which signals predict a repeat purchase. Flag nothing and ask it to group them, and an unsupervised model might suggest four clusters: monthly buyers, holiday gift buyers, discount hunters and dormant customers. Naming them and picking a plan for each is your job.
Reinforcement is the one you'll touch least. It sits behind how chatbots get fine-tuned: in a 2022 paper, OpenAI described training its InstructGPT models with reinforcement learning from human feedback.
The decision rule is simple: look at what data you have. Past customers with a known outcome (bought or didn't) point to supervised. Lots of customers and no labels point to unsupervised. And deep learning is just machine learning with many-layered neural networks, the engine under generative AI and computer vision.
Types of AI for marketing: what each one is for
By task, there are six types of AI you can actually use in marketing, and each one does a different job you can measure.
| Type | What it does | Marketing use | What to track | Tools to try |
|---|---|---|---|---|
| Generative | Creates text, images, video or voice | Drafts of posts, ads and product pages | Minutes per piece, CTR | ChatGPT, Gemini, Claude, Canva, Adobe Firefly |
| Predictive | Forecasts what's likely to happen | Bidding, audiences, customers about to churn | CPA, ROAS, churn rate | Google Analytics 4, Google Ads, Meta Advantage+, Klaviyo |
| Conversational | Understands and replies in natural language | 24/7 support and lead capture on your site or DMs | Questions resolved without a human, leads | ManyChat, Tidio, Intercom |
| Computer vision | Recognizes what's in an image | Visual search, catalog tagging, screening customer photos | Hours spent cataloging | Google Lens, Google Cloud Vision, Amazon Rekognition |
| Recommendation | Picks what to show each person | Related products, personalized emails | Average order value, cross-sells | Shopify Search & Discovery, Klaviyo, Amazon Personalize |
| Agents | Chain steps together to hit a goal | Competitor research, building reports | Hours saved, errors | ChatGPT agent mode, Zapier Agents, n8n |
Generative AI is the most used and the most misused. Use it for the first draft, never the last: our guide to AI for social media walks through a full week of content, and before you hit publish, learn how to humanize AI text so it sounds like you.
Conversational AI runs assistants and chatbots. If you're torn between a rules-based bot and an AI one, start with what is a chatbot and which kind fits your business.
Are you already using predictive AI without knowing it?
Almost certainly. If you run Smart Bidding in Google Ads or Advantage+ campaigns on Meta, a predictive AI model decides who sees your ads and how much to bid. Google Analytics 4 also calculates purchase probability, churn probability and predicted revenue for your users. According to Google Analytics Help, checked in late October 2026, you need at least 1,000 returning users who met the condition and another 1,000 who didn't over the last 28 days.
Which types of AI tools are free?
Most have a free tier. ChatGPT, Gemini, Claude and Copilot offer free plans for generative and conversational AI, Canva lets you create a limited number of AI images and Google Lens costs nothing. The predictive AI in GA4, Google Ads and Meta is included: you don't pay extra, you pay for the ads. What usually costs money is agents and chatbots handling lots of conversations a month.
Types of generative AI: text, image, video, audio and code
Generative AI splits by what it creates, and each type has its own engine, its own place in marketing and its own risk.
| Type | What it creates | Typical engine | Example tools | Always check |
|---|---|---|---|---|
| Text | Posts, emails, product descriptions | Large language model (LLM) | ChatGPT, Gemini, Claude | Invented facts and numbers |
| Image | Illustrations, backgrounds, product variants | Diffusion model | Adobe Firefly, Midjourney, Canva | Hands, text and logos |
| Video | Short clips and ads | Video diffusion model | Veo, Runway | Consistency from shot to shot |
| Audio and voice | Voiceovers and music | Speech and music synthesis models | ElevenLabs | Permission to use the voice |
| Code | Scripts, templates, automations | LLM | GitHub Copilot, Claude Code | Security before you publish |
The lines are blurring: multimodal models such as GPT and Gemini handle text, images and audio in one place. Even so, each format fails in its own way, so each gets reviewed in its own way.
Start with text and images: they're the cheapest to test and the easiest to review in a minute. Video and voice need a locked script and more rounds of fixes. Code only pays off if someone on your team can read it. Images, video and voice each have their own guide on Notablia.
How to choose types of AI tools for your business
To choose types of AI tools, start with the task that eats the most hours, match it to a type and test two options against one number.
-
Write down the task, not the tool
List the five marketing tasks that take up most of your week: answering DMs, shooting catalog photos, writing the newsletter.
-
Turn it into a verb
Create is generative, predict is predictive, answer is conversational, recognize images is computer vision, suggest is recommendation and chain steps is an agent.
-
Check the AI you already pay for
Google Ads, Meta, Canva, your CRM and your store platform already ship AI features. Turn them on before you buy anything new.
-
Test two tools on the same task
One week, same task, same instructions: results, not vibes.
-
Measure one number before and after
Minutes per piece, cost per lead or questions resolved without you. If the number doesn't improve, the tool goes.
-
Decide what you won't delegate
Prices, promises, customer data and the final review stay with you. And if you sell to Europe, check what the EU AI Act requires for chatbots and AI-generated content.
Examples of AI types working together in a small business
A single small business can run four types of AI at once, each on its own task, without hiring anyone.
Picture a coffee roaster in Austin selling online and from a small shop. Generative AI drafts the product pages and the weekly email. Meta's predictive AI spreads the ad budget. A conversational AI bot answers shipping questions in Instagram DMs. And the store's recommendation engine shows a grinder to anyone buying whole beans.
My business is [type of business] in [city] and I sell [product] to [type of customer]. These are the five marketing tasks that take me the most time: [list]. For each one, tell me which type of AI handles it (generative, predictive, conversational, computer vision, recommendation or agent), which number I should track before and after, and what I shouldn't delegate. Don't recommend specific tools yet.
Compare two [type] AI tools for this task: [task]. Search the web for their current plans and limits, tell me what each free plan includes, how they handle the data I give them and which number would show me within a week which one works. Cite the official pages you use.
Frequently asked questions about the types of AI
What are the top 5 most used AI?
There's no official ranking, but by publicly reported users the big five chat assistants are ChatGPT, Google Gemini, Meta AI, Microsoft Copilot and Claude, with Perplexity close behind. ChatGPT leads by a wide margin: OpenAI reported 800 million weekly users in October 2025. Keep in mind that the AI most people use every day is invisible, like spam filters, Netflix and Amazon recommendations, Face ID and the bidding algorithms behind every ad you see.
Which AI is better than ChatGPT?
None is better at everything, so pick by task. Claude is a favorite for long-form writing and editing, Gemini works best if you live in Google Workspace, Perplexity is built for researching with sources, and Copilot fits teams on Microsoft 365. The honest test is to give two tools the same task and the same prompt for a week.
What is the difference between AI and machine learning?
AI is the goal and machine learning is the most common way to get there. Artificial intelligence covers any system that performs tasks we associate with human intelligence. Machine learning is the technique of teaching a system by feeding it examples instead of writing rules by hand. Almost all modern AI, ChatGPT included, is built with machine learning. Older rule-based systems were AI without it.
What is computer vision?
Computer vision is the branch of AI that interprets images and video: it recognizes objects, faces, text and scenes. It powers Google Lens, Face ID and cars that read traffic signs. In marketing, you use it for visual product search, for tagging thousands of catalog photos automatically and for screening user-generated content before it goes live.
What is artificial general intelligence?
Artificial general intelligence, or AGI, would be a system that can learn and handle any intellectual task as well as a person, without being trained for each one. It doesn't exist today. Some AI labs say they are getting close and many researchers dispute it, so be skeptical of any firm date. For your business, the practical point is that every AI product you can buy is narrow AI, and you should choose it by task.
What is conversational AI?
Conversational AI is the type of AI that understands and answers people in natural language, by text or voice. It's what runs modern chatbots, voice assistants and AI customer service agents. Unlike old rule-based bots that follow a fixed menu, it can handle questions phrased in many ways. For a small business, the typical use is answering common questions and capturing leads around the clock on your website, Instagram or WhatsApp.