YouTube and TikTok rarely act on climate misinformation, even when it breaks their own rules

A joint investigation by Science Feedback and Fundación Maldita.es reported more than 300 climate misinformation videos in French and Spanish. YouTube never answered, and TikTok restricted only a small minority of them and overturned none of its decisions on appeal.

How we tested the platforms

We searched YouTube and TikTok for terms often associated with climate misinformation, based on our observations over the years. An AI model flagged 331 TikTok and 894 YouTube videos, in French and Spanish, as likely to break the platforms’ rules. Fact-checkers then reviewed the most-viewed ones and confirmed 87 on TikTok and 82 on YouTube that indeed violated the platforms’ policies on climate misinformation. We reported them through the platforms’ reporting tools, then looked for other videos from known repeat offenders and reported 87 more on TikTok and 81 more on YouTube.

On YouTube we reported only monetised videos, because demonetisation is the only enforcement pathway the platform sets out for climate misinformation. On TikTok we used the “harmful misinformation” category, under which TikTok’s own rules prohibit misinformation that denies the existence of climate change, misrepresents its causes or contradicts its established environmental impact.

YouTube provided no answer, and the videos keep earning revenue

YouTube has no content rule on climate misinformation. Since 2021 it says only that it does not allow ads on content that contradicts the scientific consensus on climate change. We did not receive a single response to our reports. YouTube’s reporting page lists only removal or age restriction as possible outcomes, so even when a report does lead to demonetisation, the person who reported it cannot tell.

This is not the first time we have tested this rule. In December 2021, weeks after it took effect, Science Feedback checked 36 videos that Climate Feedback reviews had found to contain climate misinformation. Of the 20 that fell under Google’s own definition, half were still monetised, through ads (8 videos) or merchandise integration (2).

For the present report, we checked each video’s monetisation status before and after reporting, using ad detection and third-party tools. Fewer than 5% of violative videos showed signs of demonetisation 72 hours after reporting. Some videos carried an information panel linking to sources such as the United Nations and Wikipedia while still earning ad revenue. A previous Science Feedback study found that such panels appeared on only 37% of 940 fact-checked misinformation videos, are generic, and also appear next to credible videos, so they are unlikely to inform viewers well. Under YouTube’s partner terms, creators receive 55% of net ad revenue and YouTube keeps the rest, so those views pay the platform too.

TikTok rarely applied its climate rule

TikTok is one of the few platforms with an explicit policy against climate misinformation: its own rules prohibit misinformation that denies the existence of climate change, misrepresents its causes or contradicts its established environmental impact. A platform that publishes a rule should apply it diligently, or the rule misleads the public.

In practice TikTok restricted only a small minority of the videos we reported. It applies two kinds of rules: some policies lead to a takedown, while others, such as its rule on conspiracy theories, lead to demotion, which means the video is recommended less in users’ feeds but stays online. In the cases we traced, TikTok acted under the demotion rule. Even then, TikTok left up many conspiratorial videos (for example on chemtrails and geoengineering) that made near-identical claims to videos it did act on.

Conspiracy narratives dominated our sample. Weather-manipulation and chemtrails claims drew about 70% of the views of the French TikTok videos in our sample and about 84% of the Spanish ones.

We appealed every decision we disagreed with. None of the 147 appeals changed TikTok’s decision. TikTok’s own transparency report says that automated systems handle 94.1% of violating content without human review. Where we could trust the timestamps, many replies arrived about half an hour after our report, a pattern that points to possible automated review.

Repeat offenders (accounts with three or more reported violations) also showed no visible account-level action on either platform.

A TikTok bug made it harder to quantify how fast it responds

While recording TikTok’s replies we found a bug in its reports page: once an appeal is filed, the page replaces the date of TikTok’s first response with the date of the appeal. Screenshots show the first response and the appeal carrying the same timestamp, down to the second, which cannot be right. In some cases the appeal even appears to predate the first decision. For most of the French cases we tracked, this means TikTok’s real response time cannot be established. The screenshots are in the report’s Annex I.

Appeals outside the platforms

The Digital Services Act gives users routes beyond the platforms. In Spain, Maldita took its disputes to an out-of-court dispute settlement body. It answered about 10% of Maldita’s complaints within a month. When it did rule, it recommended reversing the platform’s decision in 61% of the cases it decided, and ruled in Maldita’s favour on procedural grounds in another 16%.

In France we filed one complaint per platform with Arcom, the national Digital Services Coordinator, under Article 53. Arcom cannot judge content or force a platform to reverse a moderation decision, so the complaints focus on conduct: whether the platforms apply their own rules diligently and handle reports and appeals properly. We will update this article when we hear back.

What should change

Platforms should:

• apply the climate rules they have published diligently and consistently, as Article 14 of the DSA requires, as a rule that is published but not enforced misleads the public;
• state clearly which narratives their rules cover, including the “new climate denial”, which no longer disputes how the climate changes but produces false claims on climate solutions and climate scientists;
• make human review a real check on automated decisions, as Article 20 requires, and tell people when a report leads to demonetisation.

Regulators should:

• check that very large platforms assess and mitigate climate misinformation as a systemic risk (Articles 34 and 35);
• enforce platforms’ duty to answer people who report content, which YouTube did not do;
• require transparency on monetisation under the EU Code of Conduct on Disinformation, which both platforms have signed;
• make out-of-court bodies faster, so that people get an answer in weeks, not months.

Read the full report

The full report, Inconsistencies in climate dis/misinformation moderation on YouTube and TikTok, is available here:

This report was produced by Fundación Maldita.es and Science Feedback with the support of the European Fact-Checking Standards Network.