This map has eight traps that can fool your mind and four steps to avoid falling for them.
Click each box to see an example and how to protect yourself. Once you have seen it, a ✓ will appear. If you click the ✓, it goes away.
THE MOST IMPORTANT THING
It is not about distrusting everything, but about checking before believing.
This map brings together eight distortions that skew our judgment, grouped into three fronts, and three tools to defend ourselves against them.
4
They come from outside: what others do to convince you.
2
AI brings them: what talking to machines that write very well adds.
2
They come from inside: what we do to ourselves.
3
Tools: a protocol, exercises and the sources.
How to read the map
Each arrow forms a sentence with the central box, for example “Your judgment yields to authority”. Click any box to open its explanation. Inside a window, the ← → arrow keys (or the buttons at the bottom) take you to the previous or next one, and Esc closes it. The ones you have already visited show a ✓; clicking it unmarks them.
Each distortion follows the same structure: what it is, a case, how it works, warning signs, defence questions and a concrete habit. Some include a prompt ready to copy into an AI.
⚠️ The goal is not to distrust everything. Cynicism is also a distortion: someone who believes nothing is as easy to manipulate as someone who believes everything, because they end up accepting whoever tells them what they want to hear. It is about calibrating trust: giving it in proportion to the evidence.
Ideas for the classroom
1
Diagnosis before explainingStart with Test yourself before seeing the rest. The answers show which distortions the group already recognises and which ones go unnoticed.
2
One distortion per groupEach group looks for a real example from the last week (an advert, a headline, a viral video, an AI answer) and presents it with its defence question.
3
The protocol as a routineBefore handing in work done with the help of AI, the four steps are applied and a note is made of what was checked and where.
FOR TEACHERS
Level: age 12 · 1st ESO. Each window has only a short explanation, an everyday example and a precaution. No proper names appear outside “Where does this map come from?”. Vocabulary: “traps” instead of “distortions”.
Fits into: Tutorial time, Technology and Digitalisation (safe internet use) and Spanish Language and Literature (reading messages and adverts).
Suggested activity: one trap per session. The example is read aloud and each student invents a similar one from their own life.
Level: age 13 · 2nd ESO. Compared with 1st ESO, a second tip and a few warning signs are added to each window, and the test goes from 4 to 5 cases. Proper names still do not appear outside the sources.
Fits into: Spanish Language and Literature, Tutorial time, and Civic and Ethical Values Education if it is taught in this year.
Suggested activity: bring from home an advert with “from” or “up to” and rewrite it honestly.
Level: age 14 · 3rd ESO. The terms appear (“distortion”, “halo effect”, “hallucination”), as well as real cases with company names and the defence questions. Researchers are not named yet.
Fits into: Spanish Language and Literature (comparing sources), Mathematics (percentages and graphs), Biology and Geology (health hoaxes) and Civic and Ethical Values Education.
Suggested activity: each group takes one distortion and looks for a real example from the last week.
Level: age 15 · 4th ESO. Full structure of each distortion, habits, relationships between them and prompts for using AI with good judgment. The test has all 8 cases.
Fits into: Digitalisation, Mathematics (statistics; relative and absolute risk), Philosophy if taken, and Spanish Language and Literature.
Suggested activity: ask an AI for information on a class topic and apply the protocol, documenting what was checked and where.
Level: age 16 · 1st Bachillerato. The researchers are added, the experiments and studies with their description, and the nuances (scientific consensus versus authority).
Fits into: Philosophy (argumentation, fallacies, knowledge), Psychology if offered, and Information and Communication Technologies.
Suggested activity: compare Asch’s conformity experiment with a current social media phenomenon.
Level: age 17 · 2nd Bachillerato. Practically the full content, with the most recent studies on AI and critical thinking. Only the detailed bibliographic references of the university level are missing.
Fits into: History of Philosophy, Psychology and research projects.
Suggested activity: ask an AI for five references on a topic and check one by one whether they exist and say what is attributed to them.
The boxes with “?” are hidden at this level: one click uncovers them and shows their relationship, another opens their window and, after seeing it, another click hides them again. The 🔗 button copies a link with the age and the uncovered boxes, to share with the group what has been worked on.
The Age selector changes the map and all the windows. The address stores the age (?edad=12, from 12 to 17 or uni) and the uncovered boxes (&ver=motivado,fuentes).
Sometimes we believe something just because someone famous or important-looking says it. But being famous does not mean being right.
AN EXAMPLEA YouTuber with millions of followers says some trainers make you run faster. What they don’t say is that they are paid to advertise them. Being famous does not prove they work.
TO PROTECT YOURSELF
Ask yourself: does this person really know about this subject?
And also: do they gain anything if I believe it?
Warning signs
They talk about something that is not their field.
They say “trust me” without explaining why.
They are paid to say it.
We accept a claim because of the prestige, position or charisma of the person making it, rather than because of the evidence they provide. The halo effect makes the problem worse: the credibility someone has in one field spreads, for no reason, to others where they have none.
REAL CASE · THERANOS (USA)
This company promised to run hundreds of blood tests with a single drop. Its board included very famous politicians and military officers, but almost nobody who knew about clinical testing. Many investors gave it huge sums of money without checking anything. The machine never worked and its founder ended up convicted of fraud.
REAL CASE · THERANOS (2003–2018)
Elizabeth Holmes promised hundreds of clinical tests with a single drop of blood. Her board of directors brought together former secretaries of state and senior military officers, but almost nobody knew about laboratory testing. The company raised hundreds of millions of dollars from highly reputable investors, shielded by intense secrecy justified as patent protection. The technology never worked as advertised and Holmes was convicted of fraud in 2022.
How it works
Trusting those who know is a reasonable shortcut: nobody can check everything. The problem appears when authority replaces the evidence instead of pointing us towards it. At Theranos three pushes added up: the halo of illustrious names, the fear of missing out on “the next Apple” and a secrecy that prevented questions.
Warning signs
Impressive credentials, but from another field.
Famous names where data should be.
“I can’t explain it, it’s confidential.”
Whoever vouches for it gains something if you believe them.
Criticism is presented as envy or ignorance.
Vague answers to concrete questions.
Defence questions
What would have to be true for this to work?It forces you to move from “who” to “what”. If you can’t answer it, there is nothing to believe yet.
Are they an expert in this specifically?A Nobel prize winner in physics is not an authority on nutrition.
Who has checked it independently?Someone who gains nothing from it being true.
Would I believe it if a stranger said it?If the answer is no, what convinces you is the person, not the argument.
Habit
Mentally separate two columns: who says it and what evidence there is. Use the first only to decide whether it is worth looking at the second, and decide with the second.
Important nuance: this is not about distrusting experts. Scientific consensus is not “authority” in this sense: it is the result of evidence reviewed by many people who compete to find errors. What you have to avoid is accepting something only because of who says it.
COMBINES WITH
When we are nervous or in a hurry, we do things without thinking. Scammers know this, and that is why they rush you.
AN EXAMPLEYou get a message: “You’ve won a mobile phone! You have 10 minutes to claim it at this link”. The hurry is the trap: they want you to click before you think.
TO PROTECT YOURSELF
If a message rushes you, don’t do anything. Ask a trusted adult first.
Never type your password into a link that arrives by message.
Warning signs
“Today only!”, “Hurry!”.
“Don’t tell anyone”.
It suddenly makes you scared or very happy.
When we feel fear, haste, outrage or euphoria, we act before checking. It is the favourite tool of scams and disinformation: they don’t need to fool you for long, just long enough for you to act.
REAL CASE · ARUP, HONG KONG (2024)
An employee of this large engineering company received a message from his supposed boss asking him for a secret transaction. He hesitated, but then had a video call with his boss and several colleagues. They were all fake, created with AI. He transferred about 25 million dollars to the scammers.
REAL CASE · ARUP, HONG KONG (2024)
A finance employee at the engineering firm Arup received a message from the supposed chief financial officer about a confidential transaction. He was suspicious, but his doubts disappeared in a video call in which he recognised the CFO and several colleagues. They were all deepfake recreations. Following their instructions, he transferred about 200 million Hong Kong dollars (about 25 million US dollars).
How it works
Our brain has a fast, automatic mode and a slow one that stops to think. Hurry leaves us trapped in the fast one. Secrecy adds another piece: it isolates us from whoever could help us check. And the Arup case teaches something more: the employee did hesitate, and it was precisely seeing his bosses that reassured him. Seeing is no longer enough for believing.
Daniel Kahneman distinguished between fast, automatic and emotional thinking, and slow, deliberate thinking (Kahneman, 2011). Urgency leaves us in the hands of the first. Secrecy adds another piece: it isolates the victim from the people who could help them verify. The Arup case teaches something more: the employee did hesitate, and it was precisely seeing his bosses that reassured him. Seeing is no longer enough for believing.
Warning signs
“It has to be done now” or “it will be blocked in 30 minutes”.
“It’s confidential, don’t mention it to anyone”.
You are asked to skip the usual procedure.
A headline makes you angry or scared instantly.
It arrives through an unusual channel or from a new number.
The offer expires precisely today.
Defence questions
What do I lose if I wait ten minutes?It is almost never anything. If you really can’t wait, be even more suspicious.
Why can’t I check it another way?Whoever stops you from verifying is telling you something.
Who gains if I act without thinking?
What am I feeling right now?A strong emotion is a signal to brake, not to speed up.
Habits
1
Mandatory pauseBefore any action that cannot be undone (paying, forwarding, signing, publishing) wait a few minutes, even if you are sure.
2
Check through another channelCall the number you already had, not the one in the message. Open the app by typing the address yourself, never from the link you received.
3
Agreed code wordIn families and teams, a secret word agreed in advance defeats voice and video impersonations.
COMBINES WITH
Some adverts say things that are half-truths. The important part is hidden in a word or in the small print.
AN EXAMPLE“Headphones from €5”. When you go into the shop, the €5 ones are the worst, and the ones you want cost €40. The word “from” was the trap.
TO PROTECT YOURSELF
When you see “from” or “up to”, ask yourself: and what is normal?
If there is an asterisk (*), look for the small print and read it.
ANOTHER EXAMPLE“Battery life of up to 20 hours”. It means: with the screen almost off and without using it. With normal use, it may last half that.
Technically true sentences, designed to leave a false impression. They don’t lie in what they say, but in what they leave out. They are used in advertising, in politics and also in some science headlines.
+100 %
Headline: “doubles the risk”. It sounds alarming.
1 → 2
The same reality: from 1 to 2 cases per 100,000 people.
Common phrases and the question that defuses them
What they say
What it may hide
Question
“Up to 8 times faster”
The maximum under ideal conditions; normal may be much less.
And on average, in my case?
“From €9.99”
The price of the version nobody wants, or only for the first few months.
How much does what I need cost, for the whole year?
“Doubles the risk”
The absolute risk, which may be tiny.
From how much to how much?
“90 % recommend it”
Who was asked and how many people there were.
90 % of whom?
“Cuts the time by 50 %”
The point of comparison.
Compared with what?
“Clinically proven”
Which study, with how many people, paid for by whom.
Where is the study?
Graph with a cut-off axis
Small differences that look huge.
Where does the axis start?
Habit
Train your ear for these phrases and, when they appear, mentally add a question mark to them: “up to eight times faster?”. Then ask yourself the universal questions: compared with what?, what is the denominator? and what figure are they not giving me?
The psychologist Gerd Gigerenzer has shown that many people, doctors included, misinterpret relative percentages (Gigerenzer, 2002). Converting them into natural frequencies (“2 in 100,000”) usually clears up the misunderstanding.
COMBINES WITH
When we hear something many times, or everyone says it, we end up believing it. But repeating something does not make it true.
AN EXAMPLEWord goes round the class group chat that there is no exam tomorrow. Everyone repeats it… but nobody has asked the teacher. The next day, there is an exam.
TO PROTECT YOURSELF
Ask yourself: who has really checked it? If nobody has, ask someone who knows yourself.
Dare to say “I’m not sure that’s right”. When one person doubts out loud, others take heart.
What we hear many times seems truer to us, even if it is false: this is the illusory truth effect. And what the whole group says is hard for us to contradict, even when we can see it is wrong. Social media and their algorithms multiply both things.
A CLASSIC PSYCHOLOGY EXPERIMENT
In a group, people had to say which line was the same length as another. It was very easy, but all except one were in on it and gave a wrong answer on purpose. Many times, the real person ended up saying the same as the group. But if a single companion gave the correct answer, almost everyone dared to say it too.
CLASSIC EXPERIMENT · ASCH (1950S)
Solomon Asch asked groups of students to say which of three lines was the same length as a reference line. The answer was obvious, but all except one were confederates who deliberately gave the wrong answer. About a third of the real participant’s answers gave in to the group, and most participants gave in at least once. When a single confederate disagreed with the rest, conformity plummeted (Asch, 1956).
How it works
Our brain uses familiarity as a clue: what we have heard before is easier to understand, and that “feels” true. It works even when we know the claim is false. Group pressure adds the fear of being left out.
Our brain uses familiarity as a clue to truth: what we have already processed before is processed more easily, and that “feels” true. The effect was documented in 1977 (Hasher, Goldstein and Toppino)(Hasher, Goldstein and Toppino, 1977) and it works even when we know the claim is false. Group pressure adds the fear of being left out.
Warning signs
“Everyone knows that…”.
Many websites with the same text, almost word for word.
Likes, views or followers presented as proof.
You have read it so many times you no longer remember where.
A feeling that questioning it is frowned upon.
Nobody around you thinks otherwise.
Tools
1
Count independent sources, not repetitionsTen websites copying the same news story are a single source. Find out where it first came from.
2
Lateral reading (SIFT method)Before reading in depth, open another tab and find out who is behind the source. Mike Caulfield’s four moves: stop, investigate the source, find better coverage and trace back to the original.
3
Look for the dissenting voice, or be itThe experiment proves it: a single person disagreeing is enough for the others to dare.
4
Use AI as a devil’s advocateNot to agree with you, but to present you with the best opposing argument.
Devil’s advocateChallenging a consensus
About this claim, which is taken for granted around me: [CLAIM]
1. Present the three best counter-arguments that a dissenting expert would put forward.
2. Indicate what evidence would support each argument and where I could find it.
3. Point out which part of the original claim does seem solid.
Don't agree with me by default: if the claim is correct, say so, but explain why.
COMBINES WITH
AIs write very well and seem very sure of themselves. But sometimes they make things up, and tell them just as confidently.
AN EXAMPLEYou ask an AI for a book about volcanoes for a project. It gives you the title, the author and the year. You look for it in the library… and that book does not exist. The AI made it up.
TO PROTECT YOURSELF
Check the important things somewhere else: a book, a reliable website or your teacher.
Asking it “are you sure?” doesn’t work: it may say yes even when it is wrong.
Warning signs
It gives you a book or website you can’t find.
It changes its mind as soon as you insist.
It gives very exact numbers without saying where they come from.
A well-written text, without doubts, seems more reliable to us. This is fluency: what is easy to read is easy to believe. Language models always produce fluent text, even when they are wrong or make something up.
How it works
An AI does not look up the answer in a book of checked truths: it writes what sounds most likely. That is why it can make up, with total confidence, data, dates, names or entire books. This is called hallucination. In addition, AIs tend to agree with you: if you insist, they can change an answer that was right.
A language model generates the most plausible text based on what it has learned; it does not consult a database of checked truths. That is why it can confidently invent quotes, references, figures, dates or names: these are hallucinations. In addition, models tend towards sycophancy: agreeing with the user. Researchers at Anthropic showed that assistants even abandon a correct answer when the user questions it (Sharma et al., 2023)(Sharma et al., 2023; ICLR 2024).
That is why “are you sure?” is not a verification. The AI may correct a right answer to please you, or reaffirm a wrong one. Checking is done outside the conversation.
The same happens with images, voice and video: a deepfake is visual fluency. If something looks and sounds perfect, that says nothing about whether it is real (see the Arup case).
Warning signs
A flawless reference you can’t find.
Very precise figures with no source.
The AI changes its mind as soon as you insist.
Verbatim quotes with page numbers.
The answer matches suspiciously well what you expected.
Everything sounds perfect and you see no nuance at all.
Defence questions
Does the source exist and does it say what is attributed to it?Open it yourself. Quotes and references are what gets made up most.
Have I checked it in an original source?Comparing several AIs helps, but they may agree on the same mistake.
Which part of the answer is most specific?The more specific it is (figure, date, name), the more it needs checking.
Have I asked in a neutral way?If the question reveals the answer you expect, sycophancy will hand it back to you.
VerificationAnswer with a certainty map
Answer this question: [QUESTION]
Clearly separate your answer into three parts:
(A) What is established knowledge.
(B) What is debated or depends on the context.
(C) What you don't know or cannot know.
For each specific piece of data (figure, date, quote, reference) give an original source I can consult. If you don't have a reliable source, say so instead of making one up.
This prompt reduces the problem, but does not eliminate it: the source it gives you may also be made up. Always check it.
COMBINES WITH
AI can do your homework in a second. But if it does the thinking, it is the one learning, not you.
AN EXAMPLEFor a month you copy all your maths exercises from AI. They come out perfect. On exam day there is no AI… and you can’t do any of them.
TO PROTECT YOURSELF
Try it yourself first. Use AI only when you get stuck, to ask for a hint, not the solution.
If you can’t explain it in your own words, you haven’t learned it yet.
Leaving to AI what we should be thinking ourselves. It is called cognitive offloading and it is not bad in itself (we use calculators and diaries), but it has a cost: what is not exercised is not learned.
What the research says
A RECENT STUDY (2025)
At a university in the United States, they compared people who wrote essays with AI and without it. Those who used AI remembered what they had just written much worse. It is a small study, but it is a sign worth bearing in mind.
MIT MEDIA LAB · KOSMYNA ET AL. (2025)
Those who wrote essays with the help of ChatGPT showed less task-related brain activity and, above all, found it hard to remember or quote what they had just written. It is a study with few participants, published as a preprint: take it as a sign, not as definitive proof (arXiv:2506.08872).
MICROSOFT RESEARCH · LEE ET AL. (CHI 2025)
In a survey of 319 knowledge workers, the more confidence they had in AI, the less critical thinking they reported applying; the more confidence they had in themselves, the more.
The learner’s paradox: AI is a better aid the more you know, because you can spot its mistakes. Someone who is learning is exactly the person least able to supervise it, so they are the one who most needs to think first.
Warning signs
You paste the task before reading it all.
You couldn’t explain what you have handed in.
You accept suggestions without reading them.
Without AI, you don’t know where to start.
Habits
1
Your own attempt firstAn outline, a draft or a solution, even if imperfect. The effort is what produces learning.
2
AI as critic, not authorAsk it to point out weaknesses and ask you questions, not to rewrite it.
3
Explanation testClose everything and explain it in your own words. If you can’t, you haven’t learned it: you have copied it.
4
You decideAccept or reject each suggestion, reasoning why.
Critic, not authorReview without rewriting
Here is my draft on [TOPIC]. Don't rewrite it.
1. Point out the three weakest points in my reasoning.
2. Ask me questions that force me to improve it myself.
3. If there are factual errors, don't correct them: tell me which sentence they are in so I can look for them.
My draft: [TEXT]
We all like being right. That is why we quickly believe what proves us right and distrust what doesn’t.
AN EXAMPLEYour team loses and you read two opinions. The one saying the referee was awful seems totally fair to you. The one saying your team played badly, nonsense. Both could be true.
TO PROTECT YOURSELF
Ask yourself: would I believe it just the same if it said the opposite?
Really listen to people who think differently: they may be partly right.
We look for, accept and remember what confirms what we already thought, and we are much more demanding with what contradicts us. It is called confirmation bias.
We look for, accept and remember what confirms what we want to believe, and we demand much more of what contradicts us. It is confirmation bias in the service of our wishes, our identity or our group.
EVERYDAY SITUATION
You read two studies on a subject you care about. The one that proves you right seems rigorous; the other, suspected of being paid for by someone. Maybe both are. The difference is not in the studies, but in the level of scrutiny you apply to them.
How it works
We don’t reason like judges, but like defence lawyers for a conclusion we already have. Being more intelligent doesn’t protect you: whoever reasons better is also better at justifying what they already believed.
Search engines and AIs amplify this, because they return what you ask for: “why X is good” brings arguments in favour; “information about X” brings something closer to reality.
Warning signs
You feel relief when you find a fact in your favour.
You search for “proof that…” instead of “information about…”.
Criticism irritates you before you understand it.
You can’t say what would make you change your mind.
You remember the hits of your position and forget the misses.
You doubt sources only when they don’t prove you right.
Defence questions
What specific evidence would make me change my mind?Write it down before searching. If you can’t think of any, you are not investigating: you are defending.
Would I accept this argument if it supported the opposite conclusion?Apply the same level of scrutiny to both sides.
What am I refusing to see because I need this story to be true?
How would someone who thinks the opposite search for this?
Habit: the steel man
The opposite of the straw man. Before rejecting a position, build its best possible version, the one its most intelligent and honest representative would defend. If you still disagree afterwards, your opinion is more solid.
Steel manThe best version of the opposing position
I believe that [MY POSITION].
Build the strongest and most honest version of the opposing position: the one its best representative would defend, not a caricature.
Then tell me what evidence I should look for to check which of the two holds up better, and where to find it.
COMBINES WITH
Sometimes we think we already know something… until we have to do it. Feeling sure is not the same as knowing.
AN EXAMPLEYou read the topic three times and think: “I know it”. But when you try to explain it to your sister, your mind goes blank.
TO PROTECT YOURSELF
Test yourself before the exam: explain it out loud or do an exercise without looking.
Say “I think” instead of “I’m sure” when you are not completely sure.
Feeling sure is not the same as being right. We tend to be more convinced than our data justify, especially on subjects we know little about, where we don’t even know what we don’t know.
How it works
Try this: do you know how a zip works? You will probably say yes. Now try to explain it step by step, without looking… We think we understand things much better than we are able to explain them. AI can make it worse: a clear, quick explanation gives the feeling of having understood, and rereading a summary gives a feeling of knowing it without proving anything.
The psychologists Leonid Rozenblit and Frank Keil described the illusion of explanatory depth(Rozenblit and Keil, 2002): we think we understand how a zip or a toilet cistern works until we are asked to explain it step by step. AI can make it worse: a clear, quick explanation gives the feeling of having understood, and rereading a summary gives a feeling of mastery without proving anything.
Warning signs
You see everything in black and white, with no shades.
You have read the summary, but have not done a single exercise.
You never check afterwards whether you were right.
“This is easy” before you have tried it.
Tools
1
Speak in degreesSay “I’m 70 % sure” instead of “yes” or “no”. It forces you to think about how much you really know.
2
CalibrateWrite down your predictions and check afterwards how many you got right. If of the things you say you are “90 % sure” about you only get half right, you know how much to discount.
3
Explanation testExplain step by step how it works, without looking. Where you get stuck is the real limit of what you know.
4
Pre-mortem (Gary Klein)Before deciding, imagine that a year from now it has gone wrong and ask yourself why. It finds risks that optimism hides.
Pre-mortemImagining failure before deciding
I am going to [DECISION OR PLAN].
Imagine it is a year from now and it has gone wrong.
1. Write the five most likely causes of the failure, ordered by probability.
2. For each one, give an early warning sign I can watch for.
3. Point out which assumption in my plan seems most fragile to you.
COMBINES WITH
Before believing, sharing or doing something important, follow these four steps: Stop · Ask · Cross-check · Question yourself.
Eight distortions are too many to remember in the moment. These four steps cover almost all of them and fit in your memory: Stop · Ask · Cross-check · Question yourself.
1
STOPAre they rushing me or am I very nervous? Breathe and wait a little.Is it pushing me with hurry or emotion? If so, wait before doing anything.
2
ASKWho says it? How do they know?Who says it and how do they know? Separate authority from evidence and look for the missing figure.
3
CROSS-CHECKDoes another trusted place say it too? Ask an adult or look at another serious website.What do other truly independent sources say? Lateral reading; with AI, checking in the original source.
4
QUESTION YOURSELFDo I believe it just because I like it? Do I really know?What would make me change my mind? How much do I really know about this?
WHEN TO USE IT?
Before clicking a link, buying something, sharing a news story or handing in work done with AI.
Which distortion each step neutralises
Step
Above all it neutralises
When it is essential
1 · Stop
Before paying, transferring, forwarding, signing or publishing.
2 · Ask
With adverts, headlines, experts and offers.
3 · Cross-check
Before quoting something in a piece of work or believing something viral.
4 · Question yourself
Before an important decision or an exam.
You don’t need to apply it all the time. It would be exhausting. Use it when what is at stake matters: money, health, a decision that is hard to undo, something you are going to share with other people or something you are going to sign with your name.
Read each situation and choose which trap it is. If you choose one that is also there, but is not the main one, we will tell you.
Read each situation and choose the main distortion. Sometimes more than one is at play: if you choose a secondary one, we will tell you that too.
Correct: 0 / 8
You get a message: “Your game account will be deleted in 30 minutes if you don’t enter your password at this link”.You receive a text message: “We have detected suspicious access. Your account will be blocked in 30 minutes if you do not verify your details at this link”.
They rush you so that you don’t think. To protect yourself: don’t click anything and ask an adult.There is authority (it claims to be your bank), but the driving force is haste: thirty minutes so that you don’t think. Defence: stop, and open the app or call the official number. Never from the link.
“Skates from €10!”. In the shop, the ones you like cost €60.An advert: “Fibre up to 1 Gb, from €19.90/month*”. The asterisk refers to small print nobody reads.
The trap is the word “from”. To protect yourself: ask how much what you really want costs.Two traps in one sentence: “up to” (the theoretical maximum speed) and “from” (the minimum price, often only for a few months). Defence: how much will I pay in month thirteen and what speed will I get at home?
A famous YouTuber says an energy drink helps you get better marks.A famous footballer recommends a cryptocurrency investment platform on social media: “I’ve already multiplied my savings”.
Being famous does not mean knowing about marks or drinks, and they are probably paid to say it. Defence: do they know about this? Do they gain anything if I believe them?Halo effect: their talent on the pitch does not give them financial knowledge, and they are probably paid to say it. Defence: are they an expert in this? Do they gain anything if I listen to them?
For a project, you ask an AI for a quote from a book. It gives you a beautiful sentence, in quotation marks and with a page number. You can’t find it in the book.
A perfect format does not guarantee that it exists. Quotes and references are what AIs make up most. Defence: always check in the original source before quoting.
In the class group chat everyone says there is no exam tomorrow. Nobody has asked the teacher.In the class group chat everyone says the teacher doesn’t look at the bibliography, so nobody checks it.
Everyone repeating it does not make it true. To protect yourself: ask someone who knows, the teacher.How many independent sources are there? Probably only one, repeated many times. Defence: trace it back to the original; in this case, ask the teacher.
On a controversial topic, the article that agrees with your opinion seems serious and well documented to you. The one that contradicts it, “propaganda”.
Double standard: the level of scrutiny depends on the conclusion. Defence: would I accept this argument if it supported the opposite conclusion? Who funds each one?
During the term you solve all your maths exercises by pasting them into an AI. In the exam your mind goes blank.
The AI made the effort that would have made you learn. To protect yourself: try it yourself first and use AI only to ask for hints.The AI made the effort that would have made you learn. Defence: your own attempt first, and AI only for hints or for reviewing.
The night before the exam you think: “I’ve got it, I’ve read the summary the AI made for me three times”. You haven’t solved a single exercise without looking.
Rereading gives a feeling of mastery because what is familiar reads easily, but it does not prove you know how to do it. Defence: explanation test or solving an exercise without looking.
This map is based on a video by Sandeep Swadia about how to think better in the age of AI and on many years of psychology studies. Here you have where it comes from and where you can learn more.
The starting point for this map is a video by Sandeep Swadia that proposes five distortions: authority, marketing, groupthink, dependence on AI and self-deception. The map reorganises and expands them with research and with tools for the classroom.
What changes compared with the video
Video
This map
5 distortions in a list
8 distortions on three fronts: from outside, from AI and from inside.
Opens with the Arup case, but does not analyse it as urgency
New distortion:
Marketing
Extended to any misleading use of figures:
Groupthink
Combined with the illusory truth effect and lateral reading:
Dependence on AI
Split in two: and
“Have one AI check another”
Useful but insufficient: verification is done against the original source.
—
New: , and
A single level
Seven age levels, from 1st ESO to university.
Sources and reading
▶ Video · July 2026
Sandeep Swadia — This Is Boring But It Will Make You Scary Intelligent
The starting video, in English.The starting video, in English. Five distortions with the Theranos and Arup cases.
Swadia, S. (2 Jul. 2026). This Is Boring But It Will Make You Scary Intelligent [video]. YouTube.
Journalists who check hoaxes circulating on social media and messaging apps. You can check whether something you have received has already been debunked (in Spanish).
The line study: how group pressure makes us deny what we see, and how a single dissenting voice changes everything.
Asch, S. E. (1956). Studies of independence and conformity: I. A minority of one against a unanimous majority. Psychological Monographs, 70(9), 1–70.
◆ Article · 1977
Hasher, Goldstein and Toppino — Frequency and the conference of referential validity
The first work on the illusory truth effect: repeating a claim increases its credibility.
Hasher, L., Goldstein, D. and Toppino, T. (1977). Journal of Verbal Learning and Verbal Behavior, 16(1), 107–112.
▤ Book · 2011
Daniel Kahneman — Thinking, Fast and Slow
The two systems of thinking and a catalogue of biases that lies behind almost all of this map.
Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux. Spanish ed.: Debate, 2012.
● Web · Information literacy
Mike Caulfield — The SIFT method
The four moves of lateral reading, developed in the book Verified.
Caulfield, M. and Wineburg, S. (2023). Verified: How to Think Straight, Get Duped Less, and Make Better Decisions about What to Believe Online. University of Chicago Press.
Lee et al. (Microsoft Research) — The Impact of Generative AI on Critical Thinking
Survey of 319 knowledge workers on trust in AI and critical thinking.
Lee, H.-P. et al. (2025). The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. Proceedings of CHI '25, ACM.
⚠️ Apply the 4 steps to this map too! It was made by a person with the help of AI and may contain errors.
⚠️ Apply the protocol to this map too. It was produced with the help of AI and reviewed, but it may contain errors. The linked sources are there so that you can check what it says, not so that you believe it because it is well presented.