Which tool to choose
The choice depends on how much you want to get your hands on the result.
- You want full control and a map that stays yours forever: use a general-purpose chatbot (ChatGPT, Claude, Gemini) to generate the Mermaid code, then paste it into the Mermaid Live Editor. You edit one line of text and the map changes. No limits, no paid account for the drawing part.
- You want the map ready without seeing code: use a dedicated generator like NoteGPT, MyLens or Whimsical. You upload the document or paste the text, and it builds the draggable diagram. Faster, less customizable in the details.
- You're studying from a long PDF or a video: Mapify and NoteGPT read the file directly (and in some cases a video URL) and extract its structure. They save you the step of copying the text by hand.
How to do it
The chatbot method works everywhere and doesn't age, because the Mermaid code is a stable standard. From browser or app, the path doesn't change.
- Open the chatbot and paste the text to transform: a chapter, your notes, the summary of a lesson. If the text is long, split it into two or three submissions.
- Ask for the map with a precise request. The operational syntax:
Turn the text above into a concept map in Mermaid code.
Use the "graph TD" syntax (top to bottom).
Put the main concept at the top and connect the sub-concepts with labeled arrows
that explain the relationship (for example: "causes", "example of", "divides into").
Keep a maximum of three levels of depth. Reply with the code block only.
- Copy the entire code block you receive, from the
graph TDline to the last one. - Go to mermaid.live, delete the example code you find on the left and paste in yours. The map appears on the right in real time.
- To fix a node, you don't redraw anything: you go back to the chatbot and ask for the change ("add a branch on the economic consequences," "remove the level on the technical details"), then paste in the new code.
If you don't find a dedicated button for diagrams in the chatbot, don't look for one: the code is generated in the normal text box. The button that matters is "copy" on the code block.
A quick check to know whether the map is good: read the labels on the arrows. If they only say "connected to" or say nothing, the map is a list in disguise. A real concept map explains why two ideas are connected.
A concrete example
Marta has to study the French Revolution for an oral test. She has six pages of confused notes. She pastes them in three submissions into the chatbot and asks for the Mermaid map with the request above. She receives a diagram with "Financial crisis" at the top, connected by the "leads to" arrow to the "Convocation of the Estates-General," which in turn "results in" the "Storming of the Bastille." Marta notices that the role of the Enlightenment is missing: she goes back to the chatbot, asks to add the branch of cultural causes, pastes the code back into mermaid.live and in two minutes has the complete diagram. She looks at it to reconstruct the cause-and-effect chain out loud, which is what the professor will ask her.
When it does NOT work (and how to fix it)
If the diagram comes out misshapen or with syntax errors
It happens that the chatbot writes a label with characters Mermaid doesn't accept (parentheses, colons, accents inside the nodes). The editor shows "Syntax error." Copy the error message, paste it into the chatbot and write: "This code gives this error in Mermaid, fix it." Almost always the problem is a parenthesis or a comma in a node: the chatbot fixes it on the first try.
If the map is huge and unreadable
You gave too much text at once and the map has thirty nodes. You can't use it like that. Ask the chatbot to split it: "Divide this map into three separate maps, one for each phase." Three small diagrams can be studied, one big one can't.
If you need the map as an image to hand in
Mermaid Live has an export button (look for the download options at the top, they save to PNG or SVG). If you can't find it because the interface has changed, take a screenshot of the diagram: for handing in an assignment it's enough.
A tip from someone who really uses it
The map generated by the AI counts as a draft, not a finished product. The moment when you really learn isn't when the AI draws the diagram: it's when you correct it. You move a node because it seems in the wrong place, you rewrite a label because "leads to" is too vague and you put in "was the spark of." That correction is the studying. A map you accept without touching, you've seen, not understood.
Frequently asked questions
Can I create the map directly from a photo of my notes?
Yes. The main chatbots read images: you upload the photo of the page and ask for the map in Mermaid as if you'd pasted the text. It works better if the handwriting is legible and the photo is straight and in focus. For very messy notes it's still worth double-checking what the AI read, because a misread word changes a node.
Are AI maps good for scientific subjects too?
Yes, but the type of diagram changes. For processes (a chemical cycle, an algorithm) ask for a "flowchart" instead of a concept map: in Mermaid it's still graph TD, but the nodes become steps and the arrows become the flow. For formulas the AI can get the symbols wrong: always double-check the numbers and the superscripts against the book.
Does a map made by the AI count as "copying"?
No, if you use it as a study tool and rework it. The difference is simple: if the map helps you understand and reconstruct a topic in your own words, it's a legitimate tool, like a summary made by a smart classmate. It becomes a problem only if you hand it in as your own original work, passing off as yours an elaboration you didn't do. The diagram helps you study; the oral test and the exam you sustain with your own head.