Nature Machine Intelligence

About Nature Machine Intelligence

Nature Machine Intelligence: A Complete Profile

Nature Machine Intelligence is a monthly, peer-reviewed scientific journal published by Nature Portfolio, part of Springer Nature, covering artificial intelligence, machine learning and robotics. Launched in January 2019, it quickly became one of the most selective and most cited venues in its field. Its electronic ISSN is 2522-5839, its standard abbreviation is Nat. Mach. Intell., and its official home is nature.com/natmachintell. The journal is published in English and appears monthly. On this Journals World record the publisher was previously rendered only as a numeric code and the country as Switzerland; the verified publisher is Nature Portfolio, based in the United Kingdom — OpenAlex lists the source with country code GB and host organisation Nature Portfolio — while Springer Nature group operations also run through Switzerland and Germany, which explains why some aggregators list Springer Nature Switzerland AG. Readers should use the Nature Portfolio imprint and the ISSN above as the stable identifiers, rather than any single aggregator's corporate address line.

This profile replaces an earlier short template summary. Every figure below is attributed to the source that published it: the journal's own metrics page on nature.com for impact and speed figures, OpenAlex for open bibliometric counts retrieved directly from its source record in October 2026, the Directory of Open Access Journals (DOAJ) search API for open-access listing status, and Springer Nature's own launch announcement for the journal's history. Where a figure could not be verified from an official or primary source, this profile says so plainly instead of estimating it.

Quick Facts

History and Launch

Springer Nature announced Nature Machine Intelligence on 23 November 2017, as one of three new Nature Research titles planned for January 2019, alongside Nature Metabolism and Nature Reviews Physics. The announcement explained the timing directly: the technologies underpinning artificial intelligence and robotics had made huge leaps over the previous decade, bringing applications such as self-driving cars, AI-led medical diagnosis and intelligent digital home assistants, and the publisher wanted a dedicated Nature-branded home for that research and for discussion of its consequences. The announcement also recalled that Nature itself had published some of the field's landmark moments, including the AlphaGo program that learned the game of Go and defeated professional human players, and AlphaGo Zero, which taught itself to play from scratch and then defeated the original program. The new journal was designed to extend that track record inside a title devoted entirely to machine intelligence.

The founding chief editor, Dr Liesbeth Venema, had previously led Nature's growth in pure and applied physics, with a specific focus on artificial intelligence and robotics in the years before the launch. That appointment signalled the journal's intended character from the start: a professional-editor journal in the Nature tradition, rather than a society journal run by an external academic board. The launch was not without controversy. Parts of the machine learning community criticised the decision to launch the journal primarily as a subscription title at a time when the field's culture centred on open conference proceedings and freely shared preprints, and a boycott was discussed in the community. The journal's response, reflected in its later policies, was to allow — indeed encourage — preprints, to let authors add a link to a preprint below the abstract where all readers can see it, and to offer an optional paid open-access route for individual articles. Understanding that origin matters for authors today: preprinting a manuscript does not prejudice submission here, which is unusual among journals of comparable selectivity and is a direct legacy of the launch debate.

Since 2019 the journal has published monthly volumes through the deep learning expansion, the arrival of large language models, and the rapid spread of AI methods into biology, chemistry, medicine and the physical sciences. Its scope statement has stayed deliberately broad, because the field it covers keeps absorbing new disciplines. The journal's own description says it publishes research that underpins artificial intelligence and robotics and explores how these advances will lead to the next phase of the digital age — a framing that gives equal weight to foundational method papers and to work on what the technology does in the world.

Aims, Scope and Subject Coverage

Nature Machine Intelligence states that it is interested in the best research from across artificial intelligence, machine learning and robotics, and in the significant impact those fields have on other scientific disciplines and on many aspects of society and industry. In practice, its published content falls into several recognisable strands, and OpenAlex topic counts for the source give a useful, data-driven picture of their relative weight. The largest single topic cluster is the ethics and social impacts of AI, which reflects the journal's deliberate decision to treat governance, fairness and societal consequence as core content rather than commentary decoration. Large clusters also cover computational drug discovery methods, machine learning in materials science, machine learning in bioinformatics, neural networks and their applications, artificial intelligence in healthcare, and explainable AI. Reinforcement learning in robotics, robot manipulation and learning, soft robotics, protein structure work, single-cell and spatial transcriptomics applications, and domain adaptation and few-shot learning all appear as substantial clusters as well.

For a prospective author, that topic map carries a practical lesson. A purely incremental methods paper, however competent, competes here against work that either changes what a method can do, demonstrates machine intelligence solving a scientific problem that mattered before the model arrived, or changes how the community thinks about the technology's effects. The journal's own positioning — research that underpins AI, plus exploration of its impact — means editors routinely ask two questions of a submission: is the advance significant for machine intelligence itself, and would readers outside the immediate subfield understand why it matters? Manuscripts that answer only the first question in a narrow way are usually better placed in a specialist venue.

Subject classifications assigned by indexing services align with that breadth. Scopus-oriented listings place the journal across computer science categories including artificial intelligence, computer networks and communications, computer vision and pattern recognition, human-computer interaction and software. Those category labels explain the journal's very high citation averages: it is read simultaneously by methods researchers, by applied scientists in fields such as biology and materials, and by a policy and ethics audience, so a strong paper can accumulate citations from several communities at once.

Article Types and Editorial Model

Like other Nature Portfolio research titles, Nature Machine Intelligence is run by a team of full-time professional editors; editorial decisions are made in-house rather than by an external editorial board of practising academics. The journal publishes original research Articles, Reviews, Perspectives, Comments, News and Views and related commissioned formats familiar across the Nature family. Research Articles carry the journal's citation weight, while the review and perspective formats are used heavily for the ethics, policy and field-shaping discussions that distinguish this title from a purely technical venue. Commissioned content is normally developed with the editors, and unsolicited proposals for review-type material should be discussed with the editorial office before a full manuscript is prepared, following the standard Nature Portfolio practice.

The professional-editor model shapes the submission experience in specific ways. The first decision is an editorial triage: editors judge novelty, significance and breadth of interest before any external review, and a large share of submissions end at that stage. The journal's official metrics page reports a median of 10 days from submission to first editorial decision, which is consistent with a fast, decisive triage desk. Manuscripts that pass triage go to external peer reviewers chosen by the editors. The same metrics page reports a median of 231 days from submission to acceptance, a figure that includes revision time; authors should therefore read the fast first decision and the long overall timeline together. A quick response does not mean a quick route to publication for papers that enter revision — it means the journal separates the triage question from the review question and answers the first one quickly.

Metrics: What the Numbers Actually Say

The journal's official metrics page, maintained by Springer Nature, reports a Journal Impact Factor of 29.8 for 2025 and a five-year Journal Impact Factor of 36.0 for 2025. It also reports an Immediacy Index of 6.0 (2025), an Eigenfactor score of 0.04488 (2025), an Article Influence Score of 11.006 (2025), a Source Normalized Impact per Paper (SNIP) of 6.392 (2025) and a SCImago Journal Rank (SJR) of 6.902 (2025). The same page reports usage for 2025 of 4,113,637 downloads and 6,261 Altmetric mentions. An earlier mirror of the metrics page showed the 2024 values — Journal Impact Factor 23.9 and five-year figure 31.8 — which illustrates how quickly this journal's citation indicators have risen and why any profile quoting them must attach a year. The figure of 29.8 that appeared in this record's old template was therefore correct for 2025, even though the template gave it no year or source.

Open bibliometrics tell the same story at article level. A direct OpenAlex source query for ISSN 2522-5839, repeated through the direct source-ID record S2912241403 with identical results, returned 1,313 works, 119,268 citations, an OpenAlex h-index of 154 and a two-year mean citedness of about 24.7. Those OpenAlex figures use OpenAlex's own counting rules, which differ from both Clarivate and Scopus, so they should not be mixed with the official impact factor in a single comparison. They are useful for a different purpose: they show breadth. A journal barely seven years old holding more than a thousand indexed works and well over one hundred thousand citations has built its position on a relatively small, heavily cited article set rather than on volume.

One caution belongs in every discussion of this title's metrics. Earlier aggregator pages quoted an h-index of 118, sourced from SCImago data; the old template on this site repeated 118 without attribution. That number and the OpenAlex h-index of 154 are not in conflict — they are outputs of different databases with different coverage and citation windows. The correct habit, followed throughout this profile, is to name the database next to the number. Likewise, the template's garbled quartile field has been discarded: the journal is a first-quartile (Q1) title in its Scopus categories per SCImago-based listings, and a single unexplained rank number, as the template printed, conveys nothing reliable.

Open Access, Charges and Licensing

Nature Machine Intelligence operates a hybrid model. The default route is subscription publication at no mandatory author charge, with an optional open-access route for authors who pay an article processing charge. OpenAlex records the journal's open-access article charge in its price history, listing USD 12,690 for 2025, up from USD 12,290 in 2024 and USD 11,390 in 2021 and 2022; authors must confirm the current charge on the journal's own site before submitting, because charges change and funder agreements and institutional deals can alter what an individual author actually pays. OpenAlex classifies the source as not itself an open-access journal (is_oa false) and counts 810 of its 1,313 works as open in some form, a share that reflects the paid options, author manuscript sharing and preprint culture around the journal rather than a fully open model. A DOAJ search API query for the ISSN returns zero results, and this profile therefore states plainly that the journal is not DOAJ-listed; that is the expected position for a hybrid Nature title and is not a quality judgement in either direction.

The journal's preprint stance deserves emphasis for authors from machine learning, where conference and preprint culture dominates. Preprints are allowed and encouraged, and a link to the preprint can be shown with the article. Authors remain responsible for the usual Nature Portfolio requirements on reporting, data and code availability: the launch-era commitment that the code and data supporting a paper's main findings should be openly available remains central to how editors and reviewers assess machine learning submissions here, and reproducibility material should be prepared as part of the submission, not as an afterthought once reviews arrive.

Indexing and Discoverability

The journal is indexed in Scopus and the Web of Science ecosystem, and it appears in Inspec and in the SCImago rankings built on Scopus data. OpenAlex lists it as a core source and records listings in several national and curated journal lists. Its articles carry DOIs under the Nature Portfolio's 10.1038 prefix, in the s42256 series visible in published citations. For librarians and evaluators, the combination that matters is simple: an ISSN-verified, Scopus- and Web of Science-covered title whose citation indicators are published by its own publisher with years attached. Any third-party page quoting an impact figure for this journal without a year should be treated as stale until checked against the official metrics page.

Who Should Submit, and What Editors Look For

The strongest fit is work that does at least one of three things exceptionally well. First, a methodological advance that changes capability — a learning method, architecture, training approach or robotic system that succeeds where prior approaches demonstrably failed, with comparisons rigorous enough to survive expert review. Second, a scientific application in which machine intelligence is the instrument of a discovery that the host discipline itself values: a biological mechanism, a material, a clinical insight. Third, an analysis of consequences — fairness, safety, governance, labour, environment — that is empirical or conceptually rigorous enough to change debate rather than repeat it. Papers in the journal's most cited tradition illustrate the bar: work on interpretable models for high-stakes decisions, on explainability methods that moved from local explanations to global understanding, and on the global landscape of AI ethics guidelines all became reference points far beyond their subfields.

Practical preparation follows from that bar. State the advance in the first paragraph of the cover material in terms a neighbouring discipline can grasp. Provide code and data in a form reviewers can actually run or inspect. Compare against the strongest current alternatives, not the most convenient ones, and report failures and limitations with the same precision as successes — a journal that publishes AI ethics and safety research will not reward selective reporting in a methods paper. Expect the acceptance rate to be low; the journal does not publish an official acceptance rate, and this profile does not invent one. Authors whose work is declined at triage should treat the speed of that decision as the system working as designed: the median ten-day first decision exists so that strong but misplaced papers can move to a specialist venue quickly.

Frequently Asked Questions

Is Nature Machine Intelligence open access? It is a hybrid journal. Reading most articles requires subscription access unless the author chose the paid open-access option; a DOAJ search for its ISSN returns no listing, confirming it is not a fully open-access title.

What is its impact factor? The journal's official metrics page reports 29.8 for 2025, with a five-year figure of 36.0. Always attach the year, because the 2024 figure was 23.9.

Who publishes it? Nature Portfolio, part of Springer Nature. Some aggregators display a Springer Nature corporate entity in Switzerland; the imprint and editorial operation are Nature Portfolio's, and OpenAlex records the source's country as GB.

Can I post a preprint? Yes. Preprinting is allowed and encouraged, and a preprint link can accompany the published article.

How fast is a first decision? The official median is 10 days to first editorial decision, and 231 days from submission to acceptance for accepted papers, per the journal's metrics page.

What does it cost to publish open access? OpenAlex's price history lists USD 12,690 for 2025. Confirm the current figure and any institutional agreement coverage on the journal's site before submitting.

Verification Notes

Identity was anchored to this record's own title and slug and to the ISSN 2522-5839 carried in its stored data, then checked against the publisher's launch announcement, the journal's official metrics page and a direct OpenAlex source fetch (S2912241403), with the direct source-ID fetch repeated and returning identical counts. A slug-variant duplicate check covering a “-2” variant and related forms found no twin record for this journal on Journals World at the time of checking. The stored template's numeric publisher code and unlabelled impact and quartile figures have been replaced by the attributed values above; its h-index of 118 is a SCImago-derived figure and is reported here as such, separate from the OpenAlex h-index of 154.

Sources: Springer Nature launch announcement (23 November 2017); Nature Machine Intelligence journal metrics page, nature.com; OpenAlex source S2912241403 (direct ISSN and source-ID fetches, October 2026); DOAJ search API, ISSN 2522-5839 (zero results); SCImago-derived aggregator listings for h-index and quartile context.