INTERNATIONAL RESEARCH / PEER-REVIEWED ARTICLE

Can AI increase participation while reducing social-media quality?

In a realistic social-media experiment with 680 people, some AI tools increased participation while reducing perceived quality and authenticity. We examine spillovers to other participants.

ACADEMIC WORK REVIEWED

The impact of generative AI on social media: an experimental study

ResearchersAnders Giovanni Møller, Daniel M. Romero, David Jurgens and Luca Maria Aiello.

Scientific Reports · 17 February 2026 · 16:9376 · Peer-reviewed experimental article · Source language: English

Open the original publication Find on Google Scholar

ENGLISH REVIEW AND COMMENTARY: DIJITALPI

We explain the paper in plain English, discuss what it may mean for organisations and identify our own examples separately.

Read the explanation Go to 3 business examples

English review: 10 September 2026 · The source date appears in the citation above.

Published review

Source and editorial checks are complete. DijitalPi did not conduct a customer experiment for this paper.

LET'S READ THE RESEARCH TOGETHER

We first explain the researchers' question, method and findings. We then discuss how to interpret the results, clearly separating DijitalPi's commentary from the source.

01 / WHAT DID THE RESEARCHERS WANT TO UNDERSTAND?

How does AI writing assistance change participation and conversation quality at the same time?

Møller, Romero, Jurgens and Aiello studied different AI interventions from the perspective of both content producers and readers.

Explaining the context · DijitalPi commentary

More comments can signal activity without producing useful, original or trustworthy discussion.

AI text becomes part of the environment seen by later participants, so effects can spread beyond the person using the tool.

02 / HOW WAS THE RESEARCH CONDUCTED?

How did they test the question?

A representative US sample of 680 participants was randomized in five-person groups to a control or one of four AI conditions: chat assistance, conversation starters, draft feedback or reply suggestions.

Understanding the method · DijitalPi commentary

A custom platform simulated common social-media discussion while keeping interventions separate. Commercial ranking algorithms and long-term use were outside the experiment.

The study measured both producer behaviour and consumer perceptions of the resulting content.

03 / RESEARCH FINDINGS

Some tools increased participation while perceived quality and authenticity fell.

Some AI interventions increased participation and content volume. At the same time, users could rate discussions as lower quality or less authentic, and negative effects carried into later parts of a conversation.

Participants: 680Users in the controlled US platform experiment.
Conditions: 5One control and four kinds of AI assistance.

Source: Scientific Reports article and DOI ↗

04 / DIJITALPI'S EXPLANATION

How should we interpret these findings?

Community management should not optimise comment volume alone. Quality and trust belong on the same dashboard.

Making suggestions optional is not enough; the whole thread may be affected by resulting text.

Disclosure of AI contribution matters for authenticity and user control, and its design also needs testing.

05 / CONCLUSION AND OPEN QUESTIONS

What did we learn, and what do we still not know?

AI assistance can increase social-media participation while weakening perceived quality and authenticity. The result is limited to one experimental platform, a short period and specific interventions.

Which AI support helps Turkish brand communities without weakening voice and trust? A local, longer-running study is needed.

Return to the researchers' original publication

Scientific Reports article and DOI Authors’ preprint record

DIJITALPI'S APPLICATION COMMENTARY

What could this look like in your organisation?

We created these scenarios to make the topic concrete. They are not cases from the paper or measured client results.

EXAMPLE 01

Comment volume becomes the success measure

AI conversation starters lead more people to post, but nobody measures whether replies are useful.

Do more comments mean a better community?

Open the recommendation for example 01

Make participation and quality separate goals.

Track comment volume and length alongside independent quality ratings, repetition and willingness to continue the discussion.

EXAMPLE 02

Non-users are affected too

Only some people receive suggestions, but everyone reads the resulting text.

Does the effect stop with tool users?

Open the recommendation for example 02

Measure outcomes at conversation level.

In addition to user metrics, compare the originality, tone and later replies across whole threads. The intervention can spread through the discussion.

EXAMPLE 03

The suggestion is used without a clear origin

A user lightly edits an AI reply and publishes it; other people do not know how it was produced.

When is disclosure needed?

Open the recommendation for example 03

Explain the production role clearly.

Use a plain disclosure for text written directly or substantially by AI. Keep the option to reject the suggestion and write independently.

TRY IT WITH YOUR TEAM

Track two outcomes in one community test.

During a four-week test, measure participation and discussion quality together. Examine changes across whole threads, including people who never use the AI tool.

Explore our social-media management approach

Source and ownership

Anders Giovanni Møller, Daniel M. Romero, David Jurgens and Luca Maria Aiello.
IT University of Copenhagen; University of Michigan; Pioneer Centre for AI.

Scientific Reports · 17 February 2026 · 16:9376 · Peer-reviewed experimental article. Open the original publication · Find on Google Scholar

The academic work belongs to the researchers named above. This page contains DijitalPi's explanatory review and original business examples; it is not a full translation of the paper.

← Return to all research articles