Extended definition
Altmetrics (alternative metrics) are indicators of the attention a research output receives outside the traditional citation circuit: mentions on social media, blogs, news outlets, public-policy documents, Wikipedia, and reference managers such as Mendeley. Jason Priem and colleagues coined the term in 2010, reacting to the slowness and narrow reach of citation-based metrics. The practice consolidated around commercial aggregators, chiefly Altmetric.com, which produces the Altmetric Attention Score (AAS, shown as the colored “donut”), and Elsevier’s PlumX. Sugimoto and colleagues (2017) provided the field’s reference review, showing that altmetrics do not measure what citations measure: they capture immediate attention and dissemination, not accumulated scholarly validation. The AAS is a weighted count that assigns a different weight to each source (a news story or a policy document outweighs a tweet), so two articles with the same raw number of mentions can show very different scores.
When it applies
Altmetrics are useful for capturing dimensions of impact that citation ignores or records too late. They apply to assessing the public dissemination and social engagement of research, especially in health, climate, and policy topics, where reach beyond academia is a legitimate part of impact. They also work as an early signal: while a citation takes months or years to appear, altmetric attention accumulates within days. Thelwall and colleagues (2013) found positive, if weak, correlations between several altmetric sources and later citation counts, suggesting that early attention carries some information about future reach. In funding portfolios and institutional reporting, altmetrics complement the impact factor and the h-index by documenting reach among non-academic audiences: administrators, clinicians, journalists, and policymakers.
When it does not apply
Altmetrics do not substitute for assessment of quality or rigor. Attention is not merit: an article can accumulate a high AAS for being controversial, sensational, or simply wrong, and Bornmann (2014) warned that most altmetric indicators correlate only weakly with genuine scientific impact. They do not apply as a standalone criterion in hiring, promotion, or funding decisions; the same misuse that DORA condemns in the impact factor returns, worsened, when an attention score becomes a proxy for value. They are not comparable across fields: disciplines with a strong online presence (biomedicine, public health) generate structurally larger scores than mathematics or the humanities. And they do not apply where the data are thin: most articles have low or zero AAS, and medians near zero make the indicator noisy for the typical paper.
Applications by field
- Health and biomedicine: the highest density of altmetric attention; Twitter/X and news outlets dominate, and mentions in WHO policy signal influence on practice.
- Environmental and climate science: strong pickup in news and public policy, where the AAS captures an impact that citation is slow to reflect.
- Social sciences and humanities: more variable use; blogs and Wikipedia matter more than microblogging, and the reading horizon is longer.
- Computing and engineering: attention concentrated in repositories and technical communities; Mendeley is usually the source with the widest coverage.
Common pitfalls
The first pitfall is conflating attention with quality, treating a high AAS as proof of good work when it measures only circulation. The second is ignoring weighting and gameability: scores can be inflated by coordinated campaigns, bots, or aggressive self-promotion, and the raw number hides the real source of the attention. The third is comparing fields with different media cultures as if the score were neutral. The fourth is forgetting temporal volatility: the AAS rises fast in the first days and then plateaus, so comparing articles of different ages without normalization distorts the reading. The fifth is relying on a single aggregator: Altmetric.com and PlumX cover different sources and assign different weights, so the same article receives different metrics depending on the vendor.