Developer guide
How can I detect claims and supporting evidence in social videos?
Ask mamitsu for every claim a creator makes and the support they give for it. Your reviewers get one clean row per claim, ready to check.
Start with a post
One row per claim, with the support next to it
Send a public video from YouTube, TikTok, Instagram or X to mamitsu and ask for each claim the creator makes. mamitsu listens to the whole video and reads what appears on screen, so a number shown in a graphic or a study mentioned in passing ends up next to the claim it supports.
A schema keeps every row the same: the claim, the support the creator presents in their own words, what kind of support it is, and any context they add. When a creator gives no support, the row says so with null and "none", which is a useful signal in itself.
: "${XTRAKT_USER_API_KEY:?Set XTRAKT_USER_API_KEY to your mamitsu API key}"
curl https://usemamitsu.com/api/v1/extract \
-H "Authorization: Bearer $XTRAKT_USER_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"url": "https://www.youtube.com/watch?v=B5KLActlXxw",
"prompt": "List each claim the creator makes in this video. For every claim, give the support they present for it (a study, a regulation, a number, an expert) in their own words, and any context they add. When the creator gives no support, use null for support_presented and \"none\" for support_type.",
"schema": {
"type": "object",
"properties": {
"topic": {
"type": "string"
},
"claims": {
"type": "array",
"items": {
"type": "object",
"properties": {
"claim": {
"type": "string"
},
"support_presented": {
"type": [
"string",
"null"
]
},
"support_type": {
"type": "string",
"enum": [
"study",
"regulation",
"statistic",
"expert",
"personal experience",
"none"
]
},
"context": {
"type": [
"string",
"null"
]
}
},
"required": [
"claim",
"support_presented",
"support_type",
"context"
]
}
}
},
"required": [
"topic",
"claims"
]
}
}'A real example
Twenty claims from one video about coffee
This is the full response for a YouTube video about decaf coffee and creamers. mamitsu lists 20 claims. Nine come with support: five cite studies or reviews, three give figures and one points to an FDA limit. The other eleven are stated without support.
The context field keeps the creator's stance. The first claim, that decaf is full of dangerous chemicals, is one the creator calls a myth and sets out to answer, and the row says exactly that.
{
"extraction_id": "ext_69428f4c-2338-4134-9f9d-420886324db0",
"data": {
"topic": "Coffee Myths Debunked",
"claims": [
{
"claim": "Decaf coffee is full of dangerous chemicals.",
"support_presented": null,
"support_type": "none",
"context": "This is an internet myth that refuses to die."
},
{
"claim": "The FDA allows up to 10 parts per million of methylene chloride residue in decaf coffee.",
"support_presented": "10 parts per million",
"support_type": "regulation",
"context": "This is in reference to the methylene chloride extraction process."
},
{
"claim": "The amount of methylene chloride residue allowed by the FDA is not harmful.",
"support_presented": "toxicology reviews",
"support_type": "study",
"context": "This is in reference to the methylene chloride extraction process."
},
{
"claim": "There is no evidence that links FDA-regulated decaf to cancer or organic toxicity.",
"support_presented": null,
"support_type": "none",
"context": "This is in reference to the methylene chloride extraction process."
},
{
"claim": "Decaf coffee retains 90 to 95 percent of coffee's beneficial compounds.",
"support_presented": "studies",
"support_type": "study",
"context": "The beneficial compounds include polyphenols, chlorogenic acids, and antioxidants."
},
{
"claim": "Decaf is linked to lower liver fat, lower diabetes risk, and reduced mortality.",
"support_presented": "large cohort studies",
"support_type": "study",
"context": "These are similar associations as regular coffee."
},
{
"claim": "Both regular and decaf coffee stimulate gastrin release.",
"support_presented": null,
"support_type": "none",
"context": "This promotes stomach acid and cholecystokinin."
},
{
"claim": "Coffee does not damage the stomach lining and does not cause ulcers.",
"support_presented": null,
"support_type": "none",
"context": "If reflux is present, avoid coffee on an empty stomach."
},
{
"claim": "Many popular creamers contain phosphate additives, hydrogenated oils, artificial flavors, and 4 to 6 grams of sugar per tablespoon.",
"support_presented": "4 to 6 grams of sugar per tablespoon",
"support_type": "statistic",
"context": "This is in reference to flavored coffee creamers."
},
{
"claim": "Phosphate additives are highly absorbable and raise blood phosphorus, vascular calcification risk, and kidney disease progression risk.",
"support_presented": null,
"support_type": "none",
"context": "This is in reference to kidney patients."
},
{
"claim": "Artificial sweeteners like sucralose and acesulfame K do not raise blood sugar directly.",
"support_presented": null,
"support_type": "none",
"context": "This is in reference to artificial sweeteners."
},
{
"claim": "Artificial sweeteners may alter gut microbiome composition and may even create insulin resistance.",
"support_presented": "several studies",
"support_type": "study",
"context": "This is in reference to artificial sweeteners."
},
{
"claim": "A flavored medium latte can deliver 350 calories and 30 to 50 grams of sugar.",
"support_presented": "350 calories and 30 to 50 grams of sugar",
"support_type": "statistic",
"context": "This is in reference to syrups and whipped cream."
},
{
"claim": "Many plant-based milks contain phosphate buffers, stabilizers, and gums.",
"support_presented": null,
"support_type": "none",
"context": "This is in reference to plant-based milks."
},
{
"claim": "Inorganic phosphates are about 90 to 100 percent absorbed versus 40 to 60 percent in whole foods.",
"support_presented": "90 to 100 percent absorbed versus 40 to 60 percent in whole foods",
"support_type": "statistic",
"context": "This is in reference to non-dairy creamers and sweetened plant milks."
},
{
"claim": "There is no increased fracture risk from coffee when calcium intake is adequate in the diet.",
"support_presented": "updated systematic reviews",
"support_type": "study",
"context": "This is in reference to the myth that coffee weakens bones."
},
{
"claim": "Coffee increases urinary calcium, but only slightly.",
"support_presented": null,
"support_type": "none",
"context": "This is in reference to the myth that coffee weakens bones."
},
{
"claim": "Coffee dehydrates you.",
"support_presented": null,
"support_type": "none",
"context": "This is a myth."
},
{
"claim": "The fluid in coffee fully compensates for its mild diuretic effect.",
"support_presented": null,
"support_type": "none",
"context": "Moderate coffee intake counts towards hydration."
},
{
"claim": "Coffee is a laxative.",
"support_presented": null,
"support_type": "none",
"context": "This is partly true."
}
]
},
"source": {
"provider": "youtube",
"url": "https://www.youtube.com/watch?v=B5KLActlXxw"
}
}Review the claims
Hand reviewers a list that is ready to check
mamitsu records what the post says and the support it presents. Your reviewers then do the checking: they follow each cited study or figure back to its original source and judge it there. Starting from a structured list means reviewers read instead of rewatching.
Sort by support_type to see the unsupported claims first, and keep the source link on every row so a reviewer can go back to the video in one click.
Across many posts
Compare what different creators claim
Ask the same question about every video in a study set and store the rows together. You can then see which claims repeat across creators and which support each one cites.
Research agents can do the same through the mamitsu MCP server: share the links, ask for the claims, and get the rows back in the conversation.
Related guides