TimeTree.org is a free, publicly accessible database that compiles thousands of published scientific studies to answer one deceptively simple question: when did any two species last share a common ancestor? Type in “human” and “dog,” and the site returns a divergence time of roughly 96 million years ago, synthesized from dozens of independent molecular studies. Now in its fifth edition, the resource covers well over 100,000 species and draws on more than 4,000 published timetrees, making it one of the largest collections of evolutionary timing data ever assembled.
What the Site Actually Does
At its core, TimeTree is a knowledge base that gathers divergence-time estimates from peer-reviewed molecular clock studies and presents them in a format that anyone can use. Scientists publish papers estimating when lineages split apart by analyzing DNA or protein sequences, calibrating those estimates against the fossil record, and applying statistical models. TimeTree collects those published results, organizes them by species pair, and lets you look them up without having to track down individual papers yourself.
The site offers three main ways to explore the data. First, you can enter two organisms and get a summary of their estimated divergence time, along with a list of every published study that has addressed that split. Second, you can request a “timeline” for a single species, which traces its evolutionary history all the way back to the origin of life, showing each major branching event along the way. Third, you can generate a “timetree” for an entire group of organisms at whatever taxonomic level you want, producing a branching diagram with time estimates on every node.
The interface accepts both common names and scientific names, so you don’t need to know that your household cat is Felis catus to use it. An earlier version even had a dedicated iPhone app, though the mobile-friendly website has largely taken over that role.
How It Builds a Single Timeline from Many Studies
Different research teams using different genes, different species samples, and different statistical methods will inevitably produce somewhat different estimates for the same evolutionary split. TimeTree doesn’t just pick one study and go with it. Instead, it aggregates results across all available studies for a given node, presenting a median or mean time along with a measure of how much the individual estimates vary.
To assemble a global “super timetree” that connects all species in the database, the project uses a method called hierarchical average linking. This approach starts with the accepted tree of relationships from NCBI’s taxonomy and then resolves uncertain branching patterns based on the divergence times reported across studies. The fifth edition of the resource improved this process by prioritizing studies in which the two groups of interest are cleanly separated in the analysis, reducing a bias that previously pushed some divergence estimates further into the past than they should have been.
Quality control matters here. The curation team flags studies whose time estimates are wildly out of step with the broader literature. In the most recent build, 13 studies containing node times that differed from the consensus by more than five-fold were excluded from the global timetree calculations entirely.
Where the Numbers Come From
Every divergence time on TimeTree ultimately traces back to the molecular clock concept. DNA accumulates mutations over time, and if you can estimate the rate at which those mutations pile up, you can work backward from the genetic differences between two living species to estimate when their lineage split. The catch is that mutation rates are not constant across all lineages or all periods of history. A fruit fly lineage accumulates changes at a very different pace than an elephant lineage, and rates can speed up or slow down over geological time.
Modern studies handle this by using “relaxed” clock models that allow the rate to vary across branches of the tree. These models typically rely on fossils to anchor the timeline at key points. For example, if the oldest known fossil of a particular group is 50 million years old, that provides a minimum age for that branch. The choice of how many fossil anchors to use, and how to translate their ages into statistical constraints, can substantially affect the resulting time estimates. Indeed, research has shown that the impact of different rate-evolution models on divergence times can be as significant as the choice of fossil calibrations themselves.
Fossils are essential but imperfect anchors. A fossil tells you that a lineage existed by at least a certain date, but not when it actually originated. The oldest known fossil of a group is almost certainly younger than the group’s true origin, because fossilization is rare and discovery is patchy. This is one reason molecular estimates of divergence times often predate the oldest fossil evidence, sometimes considerably.
The Gap Between Molecules and Fossils
One of the most persistent tensions in evolutionary biology is the frequent mismatch between when molecular data say a group originated and when the first fossil of that group appears. This is not a minor bookkeeping issue. For birds, a systematic comparison found that molecular divergence estimates were, on average, more than twice the age of the oldest fossil in the corresponding groups. Somewhat counterintuitively, the gap was even larger for splits within major bird lineages than for splits between them.
This pattern is not unique to birds. Early attempts to date the diversification of complex animal life using a strict molecular clock placed the origin of major animal groups roughly 100 million years before the Cambrian explosion, the period when animal fossils first become abundant. Relaxed clock methods brought those estimates somewhat closer to the fossil record, but a significant gap remained.
The causes of this mismatch are debated. Part of the discrepancy almost certainly reflects gaps in the fossil record: organisms that live in environments where fossilization is unlikely, or that have soft bodies, will leave little trace. But some of the gap may also reflect systematic biases in molecular methods. One well-documented phenomenon is “deep root attraction,” where certain modeling choices push divergence estimates toward unreasonably old dates. This tends to happen when there is conflict between the molecular signal and the physical characteristics of fossils, and standard models struggle to reconcile the two.
For users of TimeTree, this means that the reported divergence times are best understood as informed estimates with genuine uncertainty, not as hard dates etched in stone. The site’s confidence intervals, discussed below, help convey that uncertainty, but they capture variation among studies rather than every possible source of error.
Understanding the Confidence Intervals
When you look up a divergence time on TimeTree, you’ll often see not just a single number but a range. For nodes where five or more studies have provided estimates, the site calculates a confidence interval using the spread of those estimates. Roughly speaking, it computes how tightly the individual study results cluster around the average, then reports a range that captures about 95% of them.
This is useful, but it is worth understanding what it does and does not tell you. The interval captures variation among published studies, which itself reflects differences in gene sampling, calibration choices, and statistical methods. It does not account for systematic biases that might affect all studies in the same direction, such as a universally poor fossil record for a particular group. The TimeTree team explicitly recommends that researchers review individual studies and their assumptions before relying on any time estimate for downstream work, especially when precision matters.
For nodes where fewer than five studies are available, the site simply shows the minimum and maximum reported times, which gives a rougher sense of the range. Some species pairs have been studied dozens of times, yielding tight confidence intervals, while others rest on just one or two analyses. The depth of evidence behind any given number varies enormously across the tree of life.
Who Uses It and for What
TimeTree was designed to serve three audiences: researchers, educators, and the general public. For researchers, the database is a practical starting point for any analysis that requires a dated phylogeny. If you’re studying how a trait evolved across a group of species, you need a tree with time estimates on it. Building one from scratch for every project is expensive and time-consuming. TimeTree provides a ready-made scaffold that can be refined as needed. Separate software tools have even been developed to map TimeTree’s divergence times onto independently constructed phylogenies, enabling the rapid generation of large timetrees where direct estimation would be impractical.
For educators, the timeline view is a powerful teaching tool. Tracing a single species back through geological time, node by node, gives students a visceral sense of deep evolutionary history that a static textbook diagram cannot match. The fact that both common and scientific names work makes it accessible even in introductory courses.
For anyone with casual curiosity, the site is simply fun. Wondering how closely related a shark is to a salmon, or when the ancestor of all flowering plants diverged from conifers, takes about five seconds to look up.
Real-World Applications in Research
Evolutionary timelines built from molecular data have become indispensable across biology. One major application is understanding how mass extinctions and climate shifts shaped the diversity of life. A study combining molecular timetrees for nearly 6,000 mammal species with fossil records for over 5,000 genera found that correcting molecular diversification estimates with fossil extinction data revealed a surge of new species formation during the Paleocene, the roughly ten-million-year window between the asteroid impact that ended the age of dinosaurs and the intense warming event known as the Paleocene-Eocene Thermal Maximum. That finding helps resolve a long-standing puzzle about whether mammals diversified explosively right after the dinosaurs disappeared or had been quietly splitting into new lineages well before.
Molecular timelines are also essential for tracking fast-evolving pathogens. An analysis of dengue virus evolution used molecular clock methods to estimate that the virus originated roughly 1,000 years ago, more recently than previously thought, and to date the emergence of its four distinct serotypes and their subtypes. For viruses, where the fossil record is essentially nonexistent, molecular clock estimates calibrated by the known dates of virus samples are often the only way to reconstruct evolutionary history.
These examples illustrate why a centralized, curated resource like TimeTree matters. Individual researchers working on mammals, viruses, plants, or fungi all need dated evolutionary trees, and having a common reference point makes it easier to compare findings across studies.
What TimeTree Does Not Do
It is easy to overinterpret the site if you don’t know its boundaries. TimeTree does not generate new divergence-time estimates from raw data. It is a synthesis of previously published results. If no study has ever estimated the divergence time for a particular pair of species, the site will come up empty, or it will return an estimate for the nearest higher-level group that has been studied. The coverage is extensive but uneven: well-studied groups like mammals, birds, and flowering plants have dense data, while many invertebrate, fungal, and microbial lineages are sparsely represented.
The site also does not adjudicate between competing studies. If two research teams arrive at very different dates for the same split, TimeTree reports both and includes them in its average. The user is expected to dig into the underlying studies if the disagreement matters for their purposes. This is a design choice, not a flaw: the resource aims to be comprehensive rather than editorial.
Finally, the tree topology that TimeTree uses as its backbone comes from the NCBI taxonomy, which is a classification, not a fully resolved evolutionary hypothesis. In parts of the tree of life where relationships are hotly debated, the branching pattern shown on TimeTree may not reflect the latest phylogenetic analyses. The divergence times are overlaid onto this scaffold, so if the scaffold is wrong in a particular spot, the timing information there should be treated cautiously.
How Method Choices Shift the Dates
A recurring theme in molecular dating is that apparently reasonable methodological choices can lead to substantially different answers. Different ways of modeling how mutation rates change across lineages, different approaches to combining gene data, and different calibration strategies all move the resulting estimates, sometimes by tens of millions of years for deep evolutionary splits.
One important source of variation is whether researchers analyze genes concatenated into a single large dataset or use methods that account for the fact that different genes can have different evolutionary histories. These two approaches can produce meaningfully different time estimates for the same set of species. Calibrating the clock with fossils versus calibrating it with directly measured mutation rates from living populations can also yield conflicting results, because the two types of information operate over very different timescales.
This is relevant for anyone interpreting TimeTree’s output, because the database pools results from studies using a wide range of methods. The reported confidence interval for a given node partly reflects this methodological diversity. A narrow interval might mean genuine agreement among studies, or it might mean that most studies used similar approaches. A wide interval is a more honest signal that the question is still unsettled.
Newer analytical tools continue to improve the situation. The RelTime method, for instance, was shown to produce confidence intervals with excellent coverage probabilities, meaning the true divergence time fell within the reported interval about 94% of the time in simulated tests. Advances in genomic sequencing have also helped, with studies using tens of millions of base pairs from dozens of genomes achieving substantially tighter estimates than earlier work based on a handful of genes.
Comparing TimeTree to Other Resources
TimeTree is not the only effort to build a comprehensive tree of life with time information, but it occupies a distinctive niche. The Open Tree of Life project, for instance, focuses on assembling the branching relationships among all known species but does not emphasize divergence times. Other databases, like TreeBASE, archive the raw phylogenetic trees from published studies without synthesizing them into a single unified timetree. TimeTree’s distinguishing feature is that it focuses specifically on when lineages split, aggregates estimates across studies, and presents the results through an interface designed for quick lookups.
Researchers who need a fully customized dated phylogeny for a specific set of species will typically still run their own molecular clock analyses, choosing their own genes, fossils, and statistical models. But TimeTree serves as a valuable sanity check, a starting scaffold, or a quick source of dates for exploratory analyses. Its value lies less in replacing bespoke analyses than in democratizing access to timing information that would otherwise be scattered across thousands of individual papers.
Viral and Pathogen Timelines
The molecular clock behaves differently in fast-evolving organisms like viruses than in slowly evolving ones like mammals. Viral genomes mutate so quickly that measurable evolution occurs over years or decades rather than millions of years. This makes it possible to calibrate viral molecular clocks using the known collection dates of virus samples, rather than relying on fossils.
TimeTree includes divergence estimates for viruses and other pathogens drawn from studies that exploit this rapid evolution. The dengue virus analysis mentioned earlier is a good example: by analyzing envelope gene sequences from samples collected over several decades, researchers estimated not only the overall age of the virus but also when its individual serotypes and genotypes emerged. This kind of temporal information is directly useful for epidemiology, helping scientists understand how quickly a pathogen diversifies and how its evolutionary history relates to patterns of human disease spread.
The inclusion of viruses alongside animals, plants, and fungi in a single database underscores the breadth of the molecular clock framework. Whether the organisms in question diverged a thousand years ago or a billion years ago, the underlying logic is the same: genetic differences accumulate over time, and calibrated models can convert those differences into approximate dates.
Ongoing Expansion and Data Growth
TimeTree has grown substantially with each edition. The resource more than tripled its species coverage between earlier versions and the edition described in a 2017 paper, reaching over 97,000 species drawn from more than 3,000 studies. The fifth edition, published in 2022, expanded further still, incorporating additional studies and refining the methods used to build the global super timetree.
As genomic data becomes cheaper to generate, the number of molecular clock studies published each year continues to rise, which feeds directly into TimeTree’s pipeline. Groups that were once poorly represented, such as many insect orders, marine invertebrates, and microbial eukaryotes, are gradually gaining coverage. The database is a living resource that reflects the current state of the field, so its estimates for any given node may shift as new studies are added.
For users, this means it is worth checking whether a divergence time you looked up a few years ago has been updated. A node that once rested on two studies might now have ten, and the consensus estimate may have shifted. The site’s transparency about how many studies underlie each estimate makes it straightforward to judge how stable any particular number is likely to be.