How Families Check Claims in the AI Era
Audience & Promise
This talk is for families, teenagers, and classrooms who want to navigate a media environment where not everything is what it appears to be and where the tools for creating convincing false content are more accessible than ever. By the end, you will have a working four-part framework — source, evidence, motive, uncertainty — for evaluating any claim you encounter. You will have used the Claim Check Ladder on a fictional example from start to finish. And you will have language for discussing AI-generated content, algorithmic feeds, and digital citizenship with learners of any age in a way that builds lasting habits rather than one-time awareness.
Speaker Notes by Timestamp
00:00 — Misinformation and AI images
Let's begin with a scenario. You are scrolling through a social media feed or a messaging app, and a friend has shared an image. The image shows something dramatic — a building collapsed in a city you recognize, or a public figure doing something shocking. Underneath, several people have written things like "this is unbelievable" and "I knew it." The post does not include a link to any news source. It does not say where the image came from. There is no date, no byline, no caption that says who took the photograph.
How do you know if it is real?
That question has always existed. People have been circulating misleading photographs and invented stories since long before the internet. But in the current environment, the question is harder to answer for two reasons that feed each other.
The first reason is that the technical cost of creating a convincing false image, video, or piece of text has dropped dramatically. Synthetic images that would have required hours of professional work can now be produced in seconds. A short video of a face saying words it never said is no longer a technical achievement that only sophisticated actors can produce. This is not a paranoid observation — it is a description of capability that is already widely available.
The second reason is that platforms designed to maximize engagement tend to surface emotionally activating content before contextually accurate content. A false but emotionally compelling claim will often spread faster and wider than an accurate, boring correction. This is not a conspiracy — it is a consequence of how engagement-optimized platforms work. Outrage, fear, and surprise generate more interactions than verification does.
These two reasons together mean that the question "how do I know if this is real?" now requires a deliberate practice, not just good instincts. Instincts calibrated in a world where photographs were expensive to fake and widely published stories had professional fact-checking behind them are not well-calibrated for the current environment.
This is not a reason to distrust everything. Most things most people encounter are not fabricated. The goal is not paranoia — it is calibrated skepticism, meaning the level of scrutiny you apply is proportional to the stakes and to the signals that warrant closer examination.
This talk gives your family a practice. Let's build it.
08:00 — Source, evidence, motive, uncertainty
The four-part framework is the core of everything else in this talk. Everything else — the Claim Check Ladder demo, the discussion of AI images, the algorithmic feeds section — is an application of these four parts. Learn these four, and you have the tool.
Source: Who is making this claim, and how does that person or organization know what they are saying they know? A source is not just a name or a logo. A source has a type — a first-hand witness, a journalist reporting second-hand, an aggregator repeating someone else's report, an anonymous post — and each type has a different relationship to the original events. The first question is not "do I trust this source?" It is "what kind of source is this, and how does that type typically know what it claims to know?"
For AI-generated content specifically, source evaluation requires one additional step: asking whether this content was produced by an automated process rather than a person who observed or reported something. An AI image generator does not observe reality — it produces plausible visual outputs based on patterns in its training data. That is a fundamentally different epistemic relationship to the claim being made, and treating an AI image as photographic evidence confuses the two categories.
Evidence: What specific evidence does the source provide to support the claim? Evidence is not assertion — saying something is true is not evidence that it is true. Evidence includes: corroborating records, multiple independent witnesses, verifiable documents, data with transparent methodology, or physical artifacts. A claim with no supporting evidence is not necessarily false, but it should be held lightly. A claim with multiple independent sources of evidence pointing in the same direction is stronger, though still not certainly true.
For images specifically, evidence evaluation includes asking whether the image context matches the claimed time and place. A photograph from a different event, a different country, or a different year is not evidence of what it is claimed to show, even if the image itself is real. Reverse image search tools can help identify whether an image has appeared in different contexts. The lesson here is not the tool — tools change — but the underlying question: does this image show what it claims to show, in the context claimed?
Motive: Why might the person or organization sharing this claim want me to believe it? This is not an accusation. It is a structural question about incentives. Every claim comes from somewhere, and many sources have legitimate motivations that you can understand without distrusting them. A public health announcement has a motive to change your behavior; that motive can be compatible with honesty. An advertisement has a motive to sell you something; that motive means claims should be examined more carefully. A social media post that asks you to share before the facts are verified has a motive to spread; that specific motive — urgency to share now, before checking — is a consistent signal that warrants extra scrutiny.
For AI-generated content, motive evaluation sometimes requires identifying who prompted the content and why, which may not be apparent from the content itself. The safest rule: if you cannot identify who created the content and why, treat it as unverified regardless of how compelling it looks.
Uncertainty: What is the honest level of certainty that this claim is true, given everything you know? Uncertainty is not failure. It is an accurate description of your epistemic state when the evidence does not support a definitive conclusion. The correct response to insufficient evidence is to say "I don't know" or "this is unverified," not to pick the most emotionally comfortable conclusion. Teaching learners to be comfortable with uncertainty — to hold a question open rather than forcing a premature answer — is one of the hardest and most important goals of media literacy.
The four-part framework is not a checklist to run in sequence. It is a way of thinking that you apply together, updating each dimension as you learn more. Finding out who the source is changes your evidence evaluation. Understanding the motive changes how you read the uncertainty. These four dimensions inform each other.
20:00 — Claim Check Ladder demo
The Claim Check Ladder is a structured walkthrough of the four-part framework applied to a specific, fictional claim. The ladder has seven steps. Each step has a question, a possible finding, and a decision point. You can stop at any step if you have enough information to decide — the purpose of the full ladder is to build the habit, not to spend thirty minutes on every claim you encounter.
Here is the fictional scenario we will use for this demonstration. Do not attempt to match this to any real event or real outlet — it is entirely fabricated for instructional purposes.
A messaging app notification appears. A contact has shared a screenshot of what appears to be a social media post. The screenshot shows a block of text claiming that the city's school district has announced it will move to a four-day school week starting next month. The text includes a sentence that says "officials confirmed." There is no link, no date, and the screenshot does not show the account name of the original poster.
Step 1: Extract the claim. What is specifically being claimed? In this case: the local school district has announced a change to a four-day school week, starting next month, confirmed by officials. Write the claim in one sentence before evaluating it. This step prevents scope creep — evaluating a different, broader or narrower claim than the one actually being made.
Step 2: Find the source. Where did this claim come from, and who is the original poster? In this case: the screenshot does not show the original source. We have a screenshot of a screenshot, possibly. The source cannot be identified from the content provided. Finding: source unknown.
Step 3: Check the evidence. What specific evidence does the claim include? In this case: the phrase "officials confirmed" is asserted but not sourced. There is no link to a district announcement, no meeting record, no named official. Finding: zero independent evidence provided within the claim itself.
Step 4: Compare a second source. Can you find a second independent source that corroborates this claim? In this case: check the school district's official communication channel — most districts have a website, an email notification system, or an official social media account. If the district had actually announced a schedule change effective next month, it would almost certainly appear there. If it does not appear there, that is significant negative evidence. Finding: no corroboration found.
Step 5: Name the uncertainty. Given steps 1-4, what is the honest uncertainty level? In this case: source unknown, zero evidence provided, no corroboration from the most relevant official source. Uncertainty is very high. The claim may be false, may be a misunderstanding, may be ahead of an official announcement, or may be from a different district in a different region. We do not know.
Step 6: Decide whether to share. Should you share this claim in your current state of knowledge? In this case: sharing an unverified claim about your school district's schedule would cause confusion among other families. Even if the claim turns out to be true later, sharing it in its current state spreads uncertainty, not information. The right action is to wait for verification before sharing.
Step 7: Reflect. What made this claim feel compelling, and what would have to be true for it to be credible? In this case: the claim is locally relevant (your school district), emotionally significant (a schedule change next month is actionable), and uses language of authority ("officials confirmed"). These are the features that make a claim feel credible before you check it. The reflection step asks you to identify which of these features is substance and which is style. "Officials confirmed" as asserted text is style. A link to the district's official announcement page would be substance.
The Claim Check Ladder does not take long once the habit is established. Running through all seven steps on this fictional example should take an adult about four minutes. For a learner doing it for the first time, ten to fifteen minutes. With practice, the early steps become nearly automatic, and you reach a conclusion faster.
34:00 — Algorithmic feeds and digital citizenship
The Claim Check Ladder handles individual claims. But most people do not encounter claims one at a time in a neutral environment — they encounter them inside algorithmic feeds that have already decided what to show based on what has generated engagement from people with similar behavior patterns to yours.
This matters for media literacy because an algorithmic feed creates a local information environment that feels like "what is happening" but is actually "what people like me engaged with." If your feed is full of posts about a particular controversy, that does not tell you whether the controversy is important to the world — it tells you that the controversy generated engagement in your engagement cluster.
For families and classrooms, this creates a specific teaching challenge: helping learners understand that the appearance of something on a feed is not evidence of its importance, accuracy, or prevalence. A claim can be simultaneously viral and false. A claim can be simultaneously accurate and completely absent from a feed. These are independent dimensions.
Practical habits for navigating algorithmic feeds:
Seek out the primary source when a story matters. If you see a claim about a scientific finding, the primary source is the original study, not a social media post summarizing a news article summarizing the study. Each step of that chain introduces the possibility of distortion. For claims where the stakes are high — health decisions, financial decisions, civic decisions — going to the primary source is worth the extra steps.
Notice urgency as a signal. Posts that include phrases like "share before they take this down," "spread the word before it's too late," or "the media won't cover this" are using urgency as a substitute for evidence. Genuine important information does not typically require you to bypass your own evaluation process. When urgency is used to discourage checking, that is a structural red flag regardless of whether the underlying claim is true or false.
Distinguish between information and argument. A news report that describes what happened is different from an opinion column arguing about what it means. Both have value, but they make different claims and should be evaluated differently. Teaching learners to identify the genre of what they are reading — report, analysis, opinion, satire, advertisement — is a foundational digital literacy skill.
AI-generated content in feeds: synthetic images, synthetic text, and synthetic video can be inserted into feeds without clear attribution. Current tools for detecting AI-generated content are imperfect and are likely to remain imperfect. The practical response is not to develop a reliable detector — there isn't one — but to apply source and evidence standards consistently. An image without verifiable provenance is not evidence, regardless of how realistic it looks.
Digital citizenship in this context means more than just not spreading misinformation — though that is important. It means actively contributing to a healthier information environment by applying the Claim Check Ladder before sharing, by labeling uncertainty when you share something you have not fully verified, and by engaging with corrections graciously when you are wrong. Modeling this behavior for younger learners is the most effective form of digital citizenship education.
45:00 — Classroom pack roadmap
The Claim Check Ladder material in this talk is designed to scale from a family conversation to a classroom unit. The classroom pack includes materials at different levels of complexity, designed for use in homeschool groups, co-ops, traditional classrooms, and youth organizations.
The pack includes: a printed Claim Check Ladder card for each learner with the seven steps and a fill-in format for running a claim through the process, a facilitator guide with five fictional claim scenarios at different difficulty levels (ranging from an obvious fabrication to a claim that is partially accurate but misleadingly framed), a discussion guide with open questions for after each scenario, a family version with simplified language for learners ages eight to twelve, and an extension activity that asks learners to design their own fictional claim scenario and explain which features make it compelling before checking.
The classroom pack is designed to be used without internet access if necessary — the fictional scenarios are self-contained, and the Claim Check Ladder process does not depend on real-time internet searching. This makes it useful in environments with restricted device access.
For ongoing use, the companion behavior in Koydo's media literacy track prompts learners to flag one claim they checked in the past week and share what they found. This low-frequency repetition builds the habit without turning claim-checking into an anxiety-producing exercise.
Worked Demo
This demo is tied to the media-claim-check module.
A family with a thirteen-year-old and a sixteen-year-old sits down for a thirty-minute media literacy session. The parent has prepared a fictional claim: a screenshot showing a post that claims a new study found that looking at screens for more than one hour per day causes permanent vision damage in teenagers.
The claim is fictional and constructed to include common features that make claims feel credible: it cites a "study," uses specific numbers, and targets a behavior the audience engages in.
Step 1: The family extracts the claim in one sentence: "A new study found that more than one hour daily of screen time causes permanent vision damage in teenagers."
Step 2: Source finding. The screenshot does not show which study, which researchers, or which publication. The family tries to find the original study by searching the district's recommended source tools. They find news articles citing various screen-time research but no study matching the specific claim. The thirteen-year-old notes: "It says 'a study' but doesn't say which one."
Step 3: Evidence check. The claim includes a number (one hour) and a specific outcome (permanent damage) but provides no link, no researcher names, no institution, and no journal. These specifics are evidence-style language without the actual evidence behind it.
Step 4: Second source comparison. The family checks an official health organization's publicly available information on eye health and screens. The information they find is more nuanced: it discusses eye fatigue and dry eye from extended screen use, recommends regular breaks, and does not characterize the damage as "permanent" or identify a one-hour threshold as a critical limit.
Step 5: The sixteen-year-old names the uncertainty: "The claim might be based on a real study that's being misrepresented, or it might be entirely made up. We can't tell which. But neither can we confirm the specific claim as stated."
Step 6: The family decides not to share the claim. The thirteen-year-old says: "Even if there's something true in here, what they're claiming isn't what we can verify."
Step 7: Reflection. The parent asks: "What made this feel credible?" The answers: it mentions a study (authority signal), it gives a specific number (precision signal), it's about something we care about (relevance signal). None of these features are evidence. All of them are style.
The session closes with the parent asking each person to bring one real or fictional claim to the next session that they want to run through the ladder together.
Output Assets (drafts to produce)
Claim check card — A pocket-sized or printable card with the Claim Check Ladder's seven steps in brief, single-question format. Designed for use during a real claim evaluation. Front: steps 1-4. Back: steps 5-7 plus the four-part framework (source, evidence, motive, uncertainty). Family-safe language suitable for ages ten and up.
Classroom pack — A complete facilitator-led curriculum unit including: five fictional claim scenarios at graduated difficulty, facilitator guide with discussion questions, learner fill-in worksheets for the Claim Check Ladder, a glossary of media literacy vocabulary (claim, source, evidence, motive, uncertainty, corroboration, primary source, algorithmic feed), and an extension activity for designing original fictional claims.
Family guide — A two-page parent or caregiver guide explaining the four-part framework and the Claim Check Ladder at age-appropriate levels for two ranges: ages 8-12 and ages 13 and up. Includes three conversation starters for discussing media at the dinner table and a note on modeling claim-checking behavior for children.
Public-Copy Candidate Summary (post-review)
How Families Check Claims in the AI Era teaches a practical four-part framework for evaluating any claim in a media environment where synthetic content is easy to create and algorithmic feeds prioritize engagement over accuracy. Through the Claim Check Ladder — a seven-step process for identifying source, evidence, motive, and uncertainty — families and classrooms learn to evaluate claims before sharing them, distinguish evidence from assertion, and navigate AI-generated images with calibrated skepticism rather than panic or blind trust. All examples used in this talk are entirely fictional.
Cross-Surface Links
- Koydo Homeschool Paths — The Claim Check Ladder module (media-claim-check) with printable card and worksheet
- Koydo Catalyst — Media and AI literacy track for teens and adult learners
- Koydo Mentor — Companion prompts for weekly claim-checking habit practice
- Koydo Storybooks — Narrative seeds exploring characters navigating false claims (fictional, family-appropriate settings)