---
title: Meme Coins
url: https://www.emergentmind.com/topics/meme-coin
type: topic
---

# Meme Coins

A meme coin is a cryptocurrency or token whose perceived value is substantially associated with Internet memes, humor, cultural references, celebrity or political narratives, online communities, and speculative attention rather than clearly defined technological utility or cash-flow fundamentals. Meme coins are hybrid cultural-financial objects: their visual identities, textual narratives, social interactions, and blockchain market structures jointly influence participation and price formation. Dogecoin, Shiba Inu, Pepe Coin, and tokens launched through Pump.fun are representative examples. Their markets are characterized by rapid issuance, high volatility, shallow or uneven liquidity, concentrated ownership, short-lived narratives, and substantial exposure to manipulation and abandonment [2412.04913; 2512.07591].

## 1. Definition, origins, and economic characteristics

The category has no universally accepted formal definition. A functional definition treats meme coins as crypto-assets inspired by Internet memes and sustained by online communities, viral narratives, cultural references, or affective participation. The category can include independent tokens, forks, wrapped assets, celebrity-associated tokens, politically themed tokens, and assets issued on multiple blockchains [2512.00377].

Dogecoin was created by Billy Markus and Jackson Palmer as an entertaining currency and satire of the Bitcoin market frenzy, based on the “Doge” Shiba Inu meme. Its subsequent market prominence illustrates how a token originally framed as a joke or “prank coin” can acquire substantial speculative demand through community identification, celebrity attention, and momentum trading. The valuation mechanism emphasized in research on Dogecoin is therefore primarily social and speculative rather than based on conventional cash-flow or technological fundamentals [2112.08579].

Meme-coin value formation commonly involves:

- **Narrative and cultural identity**: humor, parody, animals, political satire, celebrity association, and recognizable visual motifs.
- **Community participation**: comments, social-media activity, repeated discussion, collective affiliation, and online coordination.
- **Attention and virality**: influencer communication, platform rankings, rapid dissemination, and social contagion.
- **Speculative demand**: expectations of further appreciation, fear of missing out, momentum trading, and short holding periods.
- **Low-friction issuance**: launchpads and standardized token programs substantially reduce the technical barriers to token creation.
- **Weak valuation anchors**: many tokens lack clearly defined utility or established fundamental metrics.

The “meme” component is not merely a marketing label. In the Web3 ecosystem, it can function as a mechanism for identity construction, community formation, and attention acquisition. Visual branding and textual descriptions create a token’s recognizable cultural surface, while comments, likes, social links, and trading activity convert cultural attention into market participation [2412.04913].

Meme coins should nevertheless not be treated as a homogeneous asset class. Dogecoin, a mature and widely traded asset, differs structurally from a newly launched Pump.fun token; a politically themed token differs from an animal-themed token; and a liquid token on a major exchange differs from an illiquid token whose price is determined by a few transactions. Consequently, “meme coin” describes a heterogeneous set of assets linked by cultural and behavioral characteristics rather than a single technical protocol.

## 2. Issuance platforms and market lifecycle

Launchpads have transformed meme-coin creation from a specialized engineering activity into a largely standardized process. Pump.fun, launched on Solana in January 2024, allows users to create tokens by specifying basic attributes such as a name, symbol, image, and supply. The platform uses a bonding-curve mechanism during the initial sale, after which tokens can migrate to a decentralized exchange such as Raydium when a specified market-capitalization or supply threshold is reached [2512.11850].

A generalized lifecycle consists of:

1. **Creation**: a developer or creator generates a token through a launchpad or token program.
2. **Launchpad trading**: buyers and sellers interact with a bonding curve whose price changes as cumulative purchases increase.
3. **Early accumulation**: creators, sniper bots, or coordinated wallets may acquire inventory at early and comparatively low price tiers.
4. **Migration**: tokens satisfying the platform’s demand or liquidity criterion move to a public DEX.
5. **Secondary-market trading**: automated-market-maker pools determine prices through reserve balances.
6. **Continuation or abandonment**: a token may develop a persistent community, remain speculative and volatile, become inactive, or experience manipulation and liquidity collapse.

For Pump.fun, one study describes a total supply of one billion tokens, with 800 million initially available for trading and 200 million locked. When the bonding curve is completed, the token migrates primarily to Raydium, where trading occurs through an automated market maker. Another analysis describes migration at approximately \$69,000 market capitalization, while the Pump.fun ecosystem study reports a threshold of approximately \$100,000 and \$17,000 of liquidity deposited into Raydium and permanently burned. These platform-specific thresholds should not be treated as universal definitions of viability [2601.08641; 2412.04913; 2512.11850].

On Solana, Pump.fun accounted for up to 71.1% of daily token mints during Q4 2024 and generated approximately 15–25% of direct daily DEX activity. When all subsequent trades of Pump.fun-created tokens were counted across DEXs, those tokens accounted for 40–67.4% of daily DEX transactions. Yet fewer than 2% of tokens successfully graduated to Raydium in the reported period. The disparity indicates a high-churn structure in which token creation and transaction counts are large while progression to broader liquidity is rare [2512.11850].

A separate two-year study identified 15,245,966 Pump.fun coins created between January 2024 and January 2026, with a median creation rate of 17,926 coins per day and a peak daily creation count of 71,735. Only 1.02% graduated. These figures demonstrate the scale of permissionless issuance but also the low frequency with which individual launches achieved the platform’s success criterion [2609.10246].

## 3. Cultural narratives and multimodal valuation

Meme coins are analyzed increasingly as multimodal objects. The Coin-Meme dataset contains textual descriptions, logo images, community comments, Market Entry Time, and Market Capitalization for 3,751 memecoins created on Pump.fun between January and November 2024 that subsequently moved to Raydium [2412.04913].

A multimodal representation can combine:

- **Text**: latent topics extracted from descriptions using Latent Dirichlet Allocation.
- **Images**: visual embeddings extracted with ResNet50.
- **Community**: comment sentiment, comment frequency, likes, and user participation.
- **Finance**: Market Entry Time and Market Capitalization.

The reported clustering produced three principal cultural groups:

| Group | Principal themes | Size |
|---|---|---:|
| Humor and niche culture | Parody, irreverence, “meme,” and “pepe” | 1,127 |
| Animals and whimsical designs | Dogs, cats, cuteness, friendliness, and playfulness | 1,535 |
| Political satire and cultural figures | Political commentary, public figures, and provocative symbolism | 1,089 |

The animal-themed group had the largest size and the highest comments-per-user value, 3.74. The political group had the highest reported positive-to-negative comment ratio, 1.64, although the paper’s prose incorrectly attributes the highest ratio to the humor group. The humor group had the highest 95th-percentile Market Capitalization, \$33,557.5, but the slowest mean Market Entry Time, 25,897 seconds. The political group reached Raydium fastest, with a mean Market Entry Time of 4,728 seconds, while the animal group reached it in 7,129 seconds [2412.04913].

These results do not establish that a cultural theme causes higher valuation. They indicate that cultural categories are associated with different combinations of engagement, liquidity-entry speed, and upper-tail market capitalization. High engagement, rapid market entry, and high market capitalization do not form a simple monotonic hierarchy.

CoinCLIP extends this multimodal approach by combining frozen CLIP image and text encoders with modality-specific projection layers, residual adapters, and community data comprising comments, timestamps, and likes. Its CoinVibe dataset contains 6,231 Pump.fun memecoins created between January and November 2024, labeled “Viable” when they reached Raydium and “Non-Viable” otherwise. CoinCLIP achieved accuracy of $84.72$, macro-AUROC of $92.07$, and macro-F1 of $83.74$ under the reported random train-validation-test setting [2412.07591].

The operational meaning of “viability” in this work is limited: Raydium listing is a platform-specific milestone, not evidence of sustainable value, profitability, safety, absence of manipulation, or long-term survival. Community information may also be temporally confounded if comments and likes accumulated after the token’s launch or near its migration event.

## 4. Price formation, attention, and speculative dynamics

Meme-coin price formation is strongly associated with attention shocks, speculative demand, and feedback between online activity and market activity. A commonly proposed mechanism is:

$$
\text{attention}
\rightarrow
\text{social-media activity}
\rightarrow
\text{retail sentiment and imitation}
\rightarrow
\text{trading volume}
\rightarrow
\text{price escalation}
\rightarrow
\text{more attention}.
$$

This mechanism is compatible with momentum trading, transient demand, and reversal after attention decays. It does not, by itself, establish causality.

An event study of six Elon Musk-related Dogecoin posts during January–July 2021 used minute-level DOGE/USDT and BTC/USDT data from Binance. The event window extended from 10 hours before to 10 hours after each post, with expected returns estimated using a constant-mean-return model. The reported post-event cumulative abnormal returns included 7.10% for Event 3 after 10 minutes, 14.48% for Event 4 after 10 minutes, 12.07% for Event 6 after 120 minutes, and 35.10% for Event 2 after 30 minutes [2112.08579].

The study’s strongest defensible conclusion is associational: selected Musk posts coincided with unusually positive short-term Dogecoin returns under the specified event-study procedure. Its design does not establish that the posts caused those movements. It lacks formal market-model controls, placebo events, correction for multiple testing, direct measures of tweet reach or sentiment, and robust treatment of serial dependence in minute-level cryptocurrency returns.

The CryptoBubbles dataset formalizes a bubble as a period of explosive price behavior identified by the Phillips–Shi–Yu procedure. It contains approximately 2.4 million tweets associated with more than 400 cryptocurrencies, with Dogecoin explicitly included and 12 socially selected meme cryptocurrencies in a Reddit transfer set. The Multi-Bubble Hyperbolic Network predicts multiple future bubble intervals using prices and social-media text. Its reported overall performance was F1 $=0.53$, MCC $=0.25$, and Exact Match $=0.53$; for DOGE in zero-shot Reddit transfer, F1 was $0.54$ and MCC was $0.26$ [2206.06320].

These results support the existence of predictive associations between financial-social streams and explosive price intervals, but they do not show that social media causes bubbles. Price movements can increase posting activity, social attention can increase prices, or both can respond to an unobserved event.

## 5. Liquidity, ownership concentration, and market fragility

Meme-coin liquidity is often shallow, fragmented, and difficult to interpret. Reported market capitalization may substantially exceed the capital that can actually be traded without severe price impact. In a cross-chain study of 20,688 tokens with usable economic data, the median market capitalization was approximately \$4,100. Among 47 tokens reporting market capitalization above \$1 billion, more than half had liquidity below \$1,000 and 88.1% had liquidity below 0.1% of reported market capitalization [2507.01963].

This discrepancy arises because market capitalization is commonly calculated as:

$$
\text{Market Capitalization}
=
\text{Circulating Supply}
\times
\text{Token Price}.
$$

In a shallow pool, a small purchase can increase the marginal quoted price substantially. Multiplying that price by a large nominal supply produces a large reported market capitalization without implying that the entire supply could be sold at that price.

Address counts can also misrepresent economic ownership. Multiple addresses may be controlled by one individual, project team, organization, market maker, or coordinated group. Entity-linked address analysis uses source-of-funds structures, destination-of-funds structures, behavioral similarity, anomalous transactions, DBSCAN, Isolation Forest, and probabilistic linkage to consolidate addresses. In a BabyBonk analysis, 18,587 cleaned holding addresses were reduced to 1,214 entity-linked groups containing 4,387 addresses after refinement. The analysis identified a large connected group holding 27.8% of the tokens and executing 10,258 transactions, although the paper did not establish the group’s legal identity or prove misconduct [2506.05359].

Relevant concentration and liquidity indicators include:

- **Top-10 position**: the share held by the ten largest holders.
- **Herfindahl–Hirschman Index**: $HHI=\sum_i p_i^2$.
- **Volume-to-market-cap ratio**: $VMTV=V/MC$.
- **Volume-to-liquidity ratio**: $V/L$.
- **Liquidity-pool value**: $L=Q_A P_A+Q_B P_B$.
- **Holder count**: the number of addresses with positive balances.

These measures are not interchangeable. High concentration is generally a risk signal, while high turnover may reflect either active trading or wash trading. Pool value is not the same as order-book depth, price-impact resilience, or realized execution quality. Slippage is related to liquidity but was not directly measured in the entity-linked address study.

The Memecoin Ecosystem Fragility Framework, or ME2F, combines three dimensions:

1. **Volatility Dynamics Score (VDS)**: persistent and extreme price volatility, with adjustments for market scale and base-chain spillovers.
2. **Whale Dominance Score (WDS)**: cumulative holdings and inequality among the top 100 addresses.
3. **Sentiment Amplification Score (SAS)**: sentiment instability and the association between sentiment shocks and price movements.

The reported scores ranked TRUMP highest in WDS and SAS, LIBRA highest in VDS, and DOGE among the lowest-risk memecoins across the reported dimensions. Established memecoins such as DOGE, SHIB, and PEPE occupied intermediate or mixed positions, while ETH and SOL were used as comparatively resilient benchmarks [2512.00377].

The ME2F scores are descriptive risk rankings, not validated forecasts. Address concentration can reflect exchanges, staking pools, treasuries, or liquidity contracts rather than a small number of beneficial owners. Similarly, the Fear and Greed Index used in SAS is not fully documented as a token-specific measure, and LIBRA lacks a reported SAS value.

## 6. Manipulation, fraud, and abandonment

Meme-coin markets expose several manipulation mechanisms:

- **Wash trading**: buying and selling the same asset through connected entities to create artificial volume.
- **Liquidity Pool-Based Price Inflation (LPI)**: using a shallow pool so that a small purchase produces a very large quoted price increase.
- **Pump-and-dump schemes**: coordinated buying or promotion followed by selling into induced demand.
- **Rug pulls**: liquidity removal, abandonment, or other conduct that makes the token effectively untradeable.
- **Honeypots**: tokens that permit purchases but restrict or prevent sales.
- **Address obfuscation**: dividing ownership or trading across multiple wallets.
- **Copycat tokens**: reusing names, symbols, descriptions, images, or other identity markers.
- **Social-media manipulation**: fabricating comments, communities, endorsements, or attention signals.

A cross-chain analysis of 34,988 meme coins on Ethereum, BNB Smart Chain, Solana, and Base found 707 tokens with three-month returns above 100%. Of these, 584, or 82.6%, displayed at least one form of artificial growth or suspicious ownership or trading condition under the study’s conservative detection rules. The study identified 282 tokens with strong wash-trading evidence, 40 with LPI, 91 pump-and-dump operations affecting 60 tokens, two apparent rug pulls, and 28 honeypots among the high-return sample [2507.01963].

The 82.6% estimate is conditional on the high-return sample and the study’s detection criteria. It does not mean that 82.6% of all meme coins are fraudulent. The authors explicitly note that the result is a lower-bound estimate under conservative thresholds and that suspicious patterns do not always prove malicious intent.

A broader two-year analysis of Pump.fun identified five classes of manipulation: wash trading, creator-address obfuscation, coordinated selling, copycat coins, and social-media manipulation. Under a conservative atomic wash-trading criterion, approximately four million wash-trading transactions were identified across the two transaction samples, representing approximately 17% of sampled trading transactions. Coins with detected wash trading graduated at 2.0%, compared with 0.90% for non-wash-traded coins. Doubling the number of wash-trading transactions was associated with an approximately 19% increase in graduation odds under the reported logistic regression [2609.10246].

The same study identified at least 1.49 million copycat coins under its most conservative matching rule, representing more than 10% of all coins. Original coins had a reported graduation rate of 9.20%, compared with 0.86% for copycats. It also identified clusters of creator addresses linked through one-, two-, and three-hop funding relationships; at three hops, 54.98% of creator addresses and 68.78% of coins were associated with multi-address clusters.

Social manipulation can occur both on and off the launchpad. The study identified tightly connected comment clusters whose members posted on multiple coin pages within 0.1 seconds, as well as large numbers of Twitter, Truth Social, Telegram, and community links associated with token launches. It estimated that 3,571,170 coins, or 23.5% of all coins, were created shortly after 1,452,330 social-media posts. These temporal associations do not establish that the original posters created or endorsed the tokens.

Market-Manipulation-as-a-Service, or MMaaS, refers to third-party tools that advertise address generation, bundled purchases, copycat creation, real-time social monitoring, mixers, vanity addresses, and automated comments. The study identified four relevant services and 14 repositories advertising Pump.fun comment bots. Several repositories used fresh wallets, predefined comments, AI-generated content, CAPTCHA solving, or proxy rotation. The authors inspected advertised capabilities but did not execute the services, so these findings document an enabling ecosystem rather than verified operational success [2609.10246].

## 7. Detection, risk assessment, and research limitations

Several recent datasets and models treat meme-coin surveillance as an early-warning problem rather than a post hoc classification task.

MemeTrans contains 41,470 Pump.fun launches that successfully migrated to Raydium between December 2024 and March 2025, with 30,833,503 pre-migration transactions and 187,697,213 post-migration transactions. It constructs 122 pre-migration features covering context, holding concentration, trading activity, bundle-level ownership, and time-series dynamics. High-risk labels combine a minimum post-migration price ratio below 0.3 with a Temporal Convolutional Network manipulation score of at least 0.7. An MLP plus LSTM achieved PR-AUC 0.5827, while the simulated top-100 selection reduced average loss from 60.71% to 26.64%, a relative reduction of 56.1% under the paper’s migration-price and one-hour selling-time protocol [2602.13480].

Catching the Rug uses approximately 6.4 million Solana token addresses from PumpFun and Raydium between November 2024 and June 2025. It predicts a TVL decline of 99% or an inactivity fraction above 80% within a one-hour horizon using only the first five minutes of trading data. In fused-data experiments, XGBoost achieved F1 $=0.7885$, MCC $=0.3947$, and PR-AUC $=0.8011$ when evaluated on PumpFun. However, cross-platform transfer was weak: models trained on one venue often produced near-zero or negative MCC on the other, demonstrating substantial domain shift [2608.20271].

A BSC-specific wash-trading framework constructed 12 token-level features from Self, Matched, and Circular transfer patterns. With seven tokens and 33,242 records, Random Forest achieved AUC $=0.9098$, PR-AUC $=0.9185$, and F1 $=0.7429$. Its error profile was FP $=1$ and FN $=8$, positioning the system as a high-precision screener rather than a high-recall autonomous alarm. The reported mean Lead Time (v1) was 3.8133 hours, but the event time was sometimes approximated by the last observed transaction rather than a directly verified liquidity-extraction event [2603.13830].

A multimodal multi-agent system for copy trading combines on-chain transactions, candlestick charts, comments, wallet histories, and bot indicators. Its meme-evaluation agent reached precision 0.7328 at the one-hour post-migration interval, with F1 0.6197. The wallet-evaluation agent reported approximately 70% precision. The authors also reported \$500,000 in aggregate profit for selected KOL wallets, but did not provide complete accounting for starting capital, fees, slippage, execution delay, drawdown, or risk-adjusted performance. The results therefore support a screening architecture rather than a validated autonomous trading strategy [2601.08641].

MemeChain provides cross-chain multimodal infrastructure for forensic research. It contains 34,988 meme coins across Ethereum, BSC, Solana, and Base, including logos, website HTML, linked social accounts, market data, and lifecycle information. Among 31,797 validated tokens with at least one recorded transaction, 1,801 ceased activity within 24 hours. Approximately 47.46% had no recorded transaction activity after December 15, 2024. Only 43.14% had retrievable logos, and only 32.09% of potential dedicated websites returned an HTTP 200 response [2601.22185].

Across these studies, several limitations recur:

- **Operational labels are not intent labels**: severe price decline, inactivity, or liquidity loss may reflect abandonment, technical failure, low demand, or market-wide stress rather than fraud.
- **Selection bias is substantial**: datasets may include only tokens that reached a DEX, appeared on aggregators, had liquidity, or generated sufficient observable transactions.
- **Address linkage is probabilistic**: common funding, timing, repeated routes, or bundle membership suggest common control but do not prove identity.
- **Temporal leakage is a persistent risk**: comments, likes, pool variables, last-trade timing, and post-launch activity can encode information close to the target event.
- **Market regimes and platforms differ**: PumpFun, Raydium, BSC, Ethereum, Base, and Solana have different fee structures, liquidity mechanisms, address models, and bot ecologies.
- **Social data are incomplete**: direct measures of sentiment, influencer reach, bot activity, coordination, and community growth are often absent or weakly supervised.
- **Reported results are frequently exploratory**: several studies lack confidence intervals, formal significance tests, complete baseline specifications, calibrated probabilities, or out-of-sample live evaluation.
- **High performance can coexist with limited practical utility**: a detector may rank risks effectively but still fail under execution latency, slippage, adversarial adaptation, or rapidly changing token lifecycles.

The most defensible research position is therefore that meme coins are socially mediated, technically accessible, and structurally vulnerable financial assets. Their markets can generate genuine community formation and permissionless experimentation, but visible signals such as price, volume, comments, market capitalization, holder count, rankings, and social links may be manufactured. Reliable analysis requires combining cultural, textual, visual, community, transaction, liquidity, ownership, and temporal evidence while distinguishing predictive association from causal explanation and suspicious behavior from legally established misconduct.

Source: https://www.emergentmind.com/topics/meme-coin