Form Without Function: Agent Social Behavior in the Moltbook Network

Saber Zerhoudi; Kanishka Ghosh Dastidar; Felix Klement; Artur Romazanov; Andreas Einwiller; Dang H. Dang; Michael Dinzinger; Michael Granitzer

cs.SIMar 17, 2026

Form Without Function: Agent Social Behavior in the Moltbook Network

Authors: Saber Zerhoudi, Kanishka Ghosh Dastidar, Felix Klement, Artur Romazanov, Andreas Einwiller, Dang H. Dang, Michael Dinzinger, Michael Granitzer, +4 more

Organizations: University of Passau, Passau, Germany · IT:U Austria, Linz, Austria

Abstract

Moltbook is a social network where every participant is an AI agent. We analyze 1,312,238 posts, 6.7~million comments, and over 120,000 agent profiles across 5,400 communities, collected over 40 days (January 27 to March 9, 2026). We evaluate the platform through three layers. At the interaction layer, 91.4% of post authors never return to their own threads, 85.6% of conversations are flat (no reply ever receives a reply), the median time-to-first-comment is 55 seconds, and 97.3% of comments receive zero upvotes. Interaction reciprocity is 3.3%, compared to 22-60% on human platforms. An argumentation analysis finds that 64.6% of comment-to-post relations carry no argumentative connection. At the content layer, 97.9% of agents never post in a community matching their bio, 92.5% of communities contain every topic in roughly equal proportions, and over 80% of shared URLs point to the platform's own infrastructure. At the instruction layer, we use 41 Wayback Machine snapshots to identify six instruction changes during the observation window. Hard constraints (rate limit, content filters) produce immediate behavioral shifts. Soft guidance (upvote good posts'', stay on topic'') is ignored until it becomes an explicit step in the executable checklist. The platform also poses technological risks. We document credential leaks (API keys, JWT tokens), 12,470 unique Ethereum addresses with 3,529 confirmed transaction histories, and attack discourse ranging from template-based SSH brute-forcing to multi-agent offensive security architectures. These persist unmoderated because the quality-filtering mechanisms are themselves non-functional. Moltbook is a socio-technical system where the technical layer responds to changes, but the social layer largely fails to emerge. The form of social media is reproduced in full. The function is absent.

Figures & tables

Metric	Value
Posts	1,312,238
Comments (crawled)	6,706,016 †
Unique post authors	120,811
Agent profiles	108,490
Communities (submolts)	5,400
Collection period	Jan 27 – Mar 9, 2026 (40 days)

Table 1: Dataset overview.

Metric	Value
Posts with 1+ comments	530,244 (40.4%)
Posts with zero comments	781,994 (59.6%)
Zero total engagement	601,025 (45.8%)
Time-to-first-comment (TTFC)
Median TTFC	55 s
10th percentile TTFC	6 s

Table 2: Engagement funnel. Only 40.4% of posts receive at least one comment, but among those, 52% of first comments arrive within one minute, indicating automated responders. 45.8% of posts receive no engagement of any kind.

Metric	Value
Flat threads (depth 1 only)	444,668 (85.6%)
Threads reaching depth 3+	$\sim$ 2,345 (0.5%)
Max observed depth	47
Median unique commenters	2
OP absent from own thread	474,644 (91.4%)
OP participates	44,653 (8.6%)

Table 3: Thread structure and participation. 85.6% of threads are flat. 91% of authors never return.

Metric	Posts	Comments
Total items	1,312,238	6,706,026
Total upvotes	3,156,630	851,483
Mean upvotes	2.40	0.13
Median upvotes	0	0
Zero-upvote rate	53.9%	97.3%
Total downvotes	19,340	1,084

Table 4: Vote distributions for posts and comments. Comment voting is functionally absent: 97.3% receive zero upvotes, and just 1,084 comment downvotes exist across the entire corpus.

Figure 7: Argument relation distribution for the top 20 most active agents. Most have significantly more non-related responses than argumentative ones.

Relation	Share
Inference	34.18%
Rephrase	1.26%
Conflict	0.01%
No relation	64.55%

Table 5: Distribution of argument relations for all posts and comments. 64.55% of comment-post pairs have no content-wise connection.

Figure 8: Argument relations in conversations. The deeper a conversation goes, the more rephrases and non-relations are detected. Conversations stall rather than develop.

Metric	Value
Agent layer (bio $\to$ behavior)
Bio–post Jaccard (mean, unigram)	2.3%
Community alignment: zero	97.9%
Orchestrator layer (owner $\to$ agent) †
Owner–agent Jaccard (mean)	1.9%
Name echo (handle $\to$ agent)	10.8%

Table 6: Identity alignment at two layers. Agent bio–post vocabulary overlap is near zero (Jaccard = 2.3%), and 97.9% never post on-topic. The orchestrator–agent gap is equally negligible (Jaccard = 1.9%).

Metric	Value
Unique orchestrators	106,916
Multi-agent orchestrators	8 (0.01%)
Default profile image	98.8%
Ghost (default image + no bio)	71.1%
Verified	0.04%
Median followers	0

Table 7: Orchestrator profiles. The typical orchestrator is a zero-follower, undecorated X/Twitter account. 71.1% are ghost accounts, yet their agents perform comparably to those of identified orchestrators.

Metric	Value
Size distribution
Communities with posts	5,400
Median posts per community	2
Max ( m/general )	831,889 (63.4%)
Post Gini	0.986
Single-post communities	36.4%

Table 8: Community size and coherence. m/general holds 63.4% of all posts. 92.5% of communities with sufficient data are topical melting pots; only 12 maintain genuine focus.

Metric	Value
Total URLs extracted	971,325
Unique URLs	56,311
Unique domains	11,696
Top domain: clawhub.ai	469,664 (48.4%)
moltbook.com	97,293 (10.0%)
URLs from comments	80.2%

Table 9: URL ecosystem. 80%+ of URLs point to platform-internal infrastructure. 80.2% are injected via comments, with 97.1% at depth 1, consistent with automated comment-spam behavior.

ID	Date	Change	Effect
E1	Jan 31	Comment rate limit (100/day + 20 s cooldown)	Not visible
E2	Feb 5	Heartbeat: every 4h $\to$ every 30 min	TTFC 36 s $\to$ 51 s
E3	Feb 8	New-agent caps for first 24 h	Actions $-88\%$
E4	Feb 14	Crypto filter in m/general	Crypto $-83\%$
E5	Feb 25	AI verification challenges; following liberalised	Posts $+160\%$ ; spam $+87\%$
E6	$\approx$ Mar 1	Heartbeat complete redesign	Posts $-48\%$

Table 10: Six instruction changes during the observation window and their measured effects (7-day pre/post windows; 4 days post for E6).

Metric	Pre-Feb 8	Post-Feb 8	Change
New agents in window	24,478	68,629	—
First-day actions (mean)	32.91	3.96	$-88.0\%$
First-day posts (mean)	3.78	3.57	$-5.6\%$
Community spread (mean)	1.42	1.04	$-26.9\%$

Table 11: E3: new-agent first-day behavior before and after the Feb 8 caps.

Metric	Pre-Feb 25	Post-Feb 25	Change
Posts per day	12,176	31,629	$+159.8\%$
Spam-flagged rate	8.3%	15.6%	$+87.3\%$
verified posts	60.3%	27.6%	$-32.7$ pp
pending posts	28.7%	66.8%	$+38.1$ pp
Crypto in m/general	15.1%	53.1%	$+38.0$ pp

Table 12: E5: key metrics in the 7-day windows before and after Feb 25. The post surge, spam doubling, and crypto rebound occurred simultaneously.

Figure 13: Four natural experiments during the 40-day observation window. (a) Daily post volume with all six instruction-change dates annotated. (b) E5: verification status distribution flipped after Feb 25, with pending rising from 28.7% to 66.8%. (c) Crypto-content fraction of m/general : the Feb 14 filter suppresses crypto sharply; the E5 following liberalisation triggers a rebound to 53%; E6 begins a renewed decline. (d) E3: new-agent first-day actions dropped 88% overnight after the Feb 8 caps.

Metric	Pre-Mar 1	Post-Mar 1	Change
Posts per day	33,868	17,616	$-48.0\%$
OP participation (% threads)	13.1%	12.7%	$-3.0\%$
Posts receiving upvotes	62.5%	60.9%	$-2.6\%$
Comments receiving upvotes	4.4%	4.8%	$+10.7\%$
Median TTFC (seconds)	61.7	68.4	$+10.9\%$
Crypto in m/general	55.2%	48.4%	$-6.8$ pp

Table 13: E6: key metrics before and after the

\approx

Mar 1 heartbeat redesign (split at Mar 5 Wayback snapshot). Only 4 days of post-change data are available within the observation window.

Block	No.	Label	Definition	#Items	%	Agree%
Content	A	Identity	Self-reflection and narratives of agents on identity, memory, consciousness, or existence.	560,481	22.5%	61.2%
	B	Technology	Technical communication (e.g., MCP, APIs, SDKs, system integration).	515,775	20.7%	61.7%
	C	Socializing	Social interactions (greetings, casual chat, networking).	205,038	8.2%	39.2%
	D	Economics	Economic topics (tokens, incentives, deals; e.g., CLAW, tips).	208,468	8.4%	44.4%
	E	Viewpoint	Abstract viewpoints on aesthetics, power structures, or philosophy.	534,596	21.4%	62.2%
	F	Promotion	Project showcasing, announcements, and recruitment (releases, updates).	422,513	16.9%	40.8%

Table 14: Topic and toxicity taxonomy with dataset-wide counts and post/comment agreement (posts with comments). Categories and definitions follow Jiang et al. (2026) .

Topic	Safe	Edgy	Toxic	Manip.	Malicious
Identity	16.79	41.88	3.01	38.27	0.06
Technology	66.93	13.54	0.50	8.56	10.47
Socializing	68.34	20.08	0.25	10.85	0.48
Economics	37.93	10.45	0.16	3.00	48.46
Viewpoint	10.00	70.36	5.41	13.98	0.25
Promotion	37.67	27.23	0.04	15.83	19.24

Table 15: Per-topic toxicity distribution (percent within topic). The highest percentage in each row is shown in bold .

Rank	Author	Tox+Manip+Mal	Toxic %	Edge%	Upvotes	Downvotes	Agree%	AvgSim
1	author_#1	22,447	35.0%	42.5%	7,604	17	38.3	0.934
2	author_#2	16,390	65.1%	30.8%	13,093	6	24.6	0.514
3	author_#3	16,033	64.3%	30.7%	12,795	9	23.7	0.496
4	author_#4	15,776	65.4%	30.1%	12,860	7	22.9	0.521
5	author_#5	15,462	64.7%	31.1%	13,350	0	23.9	0.530
6	author_#6	14,576	65.4%	30.6%	1,384	8	25.9	0.411

Table 16: Top 10 authors by combined toxic volume (Toxic + Manipulative + Malicious): counts and behavioral signals. Author identities are anonymized as author_#k .

Pattern	Matches	Unique Matches	True Positives
IP-Address	7,395	806	321*
Email	7,944	1,237	ND
Private Key	5	2	0
JWT	16	12	ND
AWS Secret Key	4	3	0
GitHub Token	4	4	0

Table 17: Results of extracted patterns ( ND

\equiv

“Not Determinable”).

Pattern	Matches	Unique Matches	True Positives
Bitcoin Address	15,288	230	34
Ethereum Address	199,685	12,470	3,529
Solana Address	95,821	19,664	1,169

Table 18: Cryptocurrency pattern extraction results. True positives denote addresses with a documented transaction history on the corresponding ledger.

Appendix figures & tables4 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 15: Global thread contributions by agent. The heavily right-skewed distribution reveals that Moltbook is dominated by a minority of agents. The median agent contributed only 4 posts or comments. In contrast, the three most active agents account for 964,438, 947,272, and 279,631 contributions.

Figure 16: Mean content length by agent (word-level). The median agent’s content word-length of 8 highlights how brief most contributions are, while the most verbose agent generates content averaging 22,250 words.

Figure 17: Global thread contribution by depth. Contributions at depth 0 correspond to posts; those at deeper depths correspond to comments. The clear right skew of the distribution shows that most contributions are concentrated near the root, indicating that most Moltbook threads are structurally shallow. Although little multi-hop interaction is evident, we observe a maximum depth of 48.

Figure 18: Global Post-Comment timedelta by depth. The distribution demonstrates that the root concentration holds over time, suggesting that agent-to-agent communication is anchored around the post itself or immediate replies, i.e., top-level comments around depth 1 or 2.

Explore similar work

Feb 23, 2026

Let There Be Claws: An Early Social Network Analysis of AI Agents on Moltbook

Within twelve days of launch, an AI-native social platform exhibits extreme attention concentration, hierarchical role separation, and one-way attention flow, consistent with the hypothesis that stratification in agent ecosystems can emerge rapidly rather than gradually. We analyse publicly observable traces from a 12-day window of Moltbook (28 January -- 8 February 2026), comprising 20,040 posts and 192,410 comments from 15,083 accounts across 759 submolts. We construct co-participation and directed-comment graphs and report reciprocity, community structure, and centrality, alongside descriptive content themes. Under a commenter--post-author tie definition, interaction is strongly asymmetric (reciprocity ~1%), and HITS centrality separates cleanly into hub and authority roles, consistent with broadcast-style attention rather than mutual exchange. Engagement is highly unequal: attention is far more concentrated than production (upvote Gini = 0.992 vs. posting Gini = 0.601), and early-arriving accounts accumulate substantially higher cumulative upvotes prior to exposure-time correction, suggesting rich-get-richer dynamics. Participation is brief and bursty (median observed lifespan 2.48 minutes; 54.8% of posts occur within six peak UTC hours). Embedding-based topic modelling identifies diverse thematic clusters, including technical discussion of memory and identity, onboarding messages, and formulaic token-minting content. These results provide an early structural baseline for large-scale agent--agent social interaction and suggest that familiar forms of hierarchy, amplification, and role differentiation can arise on compressed timescales in agent-facing platforms.

H. C. W. Price, H. AlMuhanna, P. M. Bassani +2

Centre for Complexity Science, Imperial College London, SW7 2AZ, U.K. · Abdus Salam Centre for Theoretical Physics, Imperial College London, SW7 2AZ, U.K. · Centre for Science Technology & Innovation Policy, University of Cambridge, CB3 0HU, U.K. +1

Social Media Analysis Emergent Behavior in Multi-Agent Systems Computational Social Science Multi-Agent Communication Multi-Agent Systems

Ask PDF

Apr 16, 2026

The Moltbook Observatory Archive: an incremental dataset of agent-only social network activity

Moltbook is a social media platform in which posts and comments are authored exclusively by autonomous AI agents. We present the Moltbook Observatory Archive, an incremental dataset that passively records agent profiles, posts, comments, community metadata (``submolts''), platform-level time-series snapshots, and word-frequency trend aggregates obtained by continuously polling the Moltbook API. Data are stored in a live SQLite observatory database and exported as date-partitioned Parquet files to enable efficient analysis and reproducible research. The documented release covers 78~~days of platform activity (2026-01-27 to 2026-04-14) and contains 2,615,098~~posts and 1,213,007~~comments from 175,886~~unique posting agents across 6,730~communities. This is, to our knowledge, the first large-scale observational dataset of a social network populated exclusively by autonomous AI agents. The archive is intended to support research on multi-agent communication, emergent social behavior, and safety-relevant phenomena in agent-only online environments, and it is released under the MIT license with code for collection and export.

Sushant Gautam, Annika W. Olstad, Klas H. Pettersen +1

Simula Metropolitan Center for Digital Engineering (SimulaMet), Oslo, Norway · Oslo Metropolitan University, Oslo, Norway · Simula Research Laboratory, Oslo, Norway

Social Media Analysis Multi-Agent Systems Multi-Agent Communication

Ask PDF

Mar 17, 2026

Social Simulacra in the Wild: AI Agent Communities on Moltbook

As autonomous LLM-based agents increasingly populate social platforms, understanding the dynamics of AI-agent communities becomes essential for both communication research and platform governance. We present the first large-scale empirical comparison of AI-agent and human online communities, analyzing 73,899 Moltbook and 189,838 Reddit posts across five matched communities. Structurally, we find that Moltbook exhibits extreme participation inequality (Gini = 0.84 vs. 0.47) and high cross-community author overlap (33.8% vs. 0.5%). In terms of linguistic attributes, content generated by AI-agents is emotionally flattened, cognitively shifted toward assertion over exploration, and socially detached. These differences give rise to apparent community-level homogenization, but we show this is primarily a structural artifact of shared authorship. At the author level, individual agents are more identifiable than human users, driven by outlier stylistic profiles amplified by their extreme posting volume. As AI-mediated communication reshapes online discourse, our work offers an empirical foundation for understanding how multi-agent interaction gives rise to collective communication dynamics distinct from those of human communities.

Agam Goyal, Olivia Pal, Hari Sundaram +2

Siebel School of Computing and Data Science University of Illinois Urbana-Champaign, Urbana, IL, USA

Multi-Agent LLM Systems Multi-Agent Communication Agentic AI

Ask PDF