History & Social Studies: How AI can best tutor this subject

What you'll be able to do: Use AI for the parts of history it's genuinely strong at (argument, interrogation, perspective) while treating everything it says about facts as unverified.

The subject where the tool is most dangerous

State this first, because everything else depends on it.

AI produces fluent, plausible, invented historical facts. Dates that are close and wrong. Quotations in the right style attributed to the wrong person. Documents described convincingly that never existed. Statistics with the shape of real statistics.

In most subjects that's a nuisance you catch later. In history, the evidence is the argument. A fabricated quotation in your essay isn't an error, it undermines the entire piece, and if it reaches a bibliography it's an academic integrity problem regardless of intent.

So the working rule for this subject is unusually strict:

> Use it for reasoning. Never use it as a source.

Every factual claim you will actually use needs a real source you have seen yourself. That isn't caution; it's the discipline's basic standard applied to a new tool.

What it is genuinely good at here

Once the sourcing question is settled, the strengths are substantial, and they're all on the argument side, which is where the marks are anyway.

Building the strongest case against your interpretation. Historiographical engagement requires stating the opposing view at its best, and students meet opposing views already weakened. A tool with no stake in the outcome that has read both traditions is genuinely useful here, and this is the single biggest band-raiser in the subject.

Interrogating a source you have in front of you. Not summarising it, asking the seven questions. Who made it, for whom, what does it take for granted, what is it arguing against, what's absent. It generates the questions; you answer them against the text.

Testing your causal reasoning. Classifying links as causal, enabling or merely sequential, and running the removal test. This is the direct fix for "descriptive" and it's mechanical enough to do well.

Surfacing your unstated premises. "Hardship causes revolt" is a claim the record contradicts, and it's invisible until named.

Multi-actor perspectives, with the invention marked.

Structural critique of your essay. What is my argument, in one sentence, as it appears on the page, not what I meant? The gap between those is usually the essay's real problem, and it's the thing you cannot see yourself.

Restricting you to what was knowable. Playing a period where it refuses to tell you what happened next is an excellent antidote to hindsight.

What it is bad at here, specifically

Facts. As above. Dates, figures, quotations, document contents, who said what to whom.

Citations. It will produce well-formatted references to work that doesn't exist, or that exists and says something else. Every reference gets chased.

Historiography as a factual matter. It's reasonable on the broad shape of debates and unreliable on who argued what. Verify attributions of position, not just of quotation.

Knowing which tradition it's speaking from. It will present one historiographical account in a neutral register, exactly as a textbook does. Ask explicitly: whose account is this, and who disagrees?

Anachronism. It slides easily into modern categories and sensibilities, which is the thing history teaching spends years removing.

Sitting with ambiguity. History's honest answer is often "we don't know" or "the evidence is thin and contested". It tends to produce a coherent account regardless.

The shape of a good session

1. Map the causation before writing, classify links, run the removal test, separate triggers from preconditions.
2. Interrogate your sources with questions generated by the tool and answered by you against the text.
3. Steelman the interpretation you reject and answer its single strongest point.
4. Write. Then reverse-outline to find what you actually argued.
5. Verify every factual claim against a real source before it enters the essay.

Notice that the tool is used at steps 1-4 and explicitly excluded at step 5.

The instruction to set

For this session you are tutoring me in history. Rules, never state a date, figure, quotation or document's contents as fact; if you refer to one, mark it as something I must verify and tell me where to look. Never give me a citation I can't check. When you describe an interpretation, say which historians hold it and that I should verify the attribution. Don't summarise sources I give you. ask me questions about them. And tell me when the honest answer is that the evidence is thin.

That last clause is worth adding everywhere and matters most here.

The failure that looks like success

A fluent, well-structured, confident essay containing three facts that aren't true and one quotation that was never said.

It reads better than your usual work. Every claim is specific. The argument flows. And it cannot survive a marker who checks one reference, and increasingly they check.

The tell: you can't say, for each factual claim in your essay, where you saw it. If the answer for any of them is "it told me", that claim isn't yours yet.

What it cannot replace

The archive, and the experience of finding that the record doesn't answer your question. History is substantially the discipline of working out what can be known from what survives, and a tool that always produces an answer teaches the opposite lesson.

Where to go next

01 Cause-consequence mapping is the direct fix for the word markers keep writing. 02 Primary source interrogation is what turns a document into evidence. 03 Steelmanning is the one that moves bands.