A turtle, according to the scientist who named it
In 1948, a joint Soviet-Mongolian paleontological expedition working the Nemegt Formation in Mongolia's Gobi Desert recovered three partial claw bones from a site labelled Quarry V, found near the skeleton of a large theropod along with a metacarpal fragment and several rib fragments. Cataloged as specimen PIN 551-483, the bones sat unexplained until Russian paleontologist Evgeny Maleev described them in 1954. Working from little more than the claws, Maleev concluded he was looking at a colossal, 4.5-metre turtle-like reptile that used its oversized hand claws to harvest seaweed from the sea floor.
Maleev named the new genus and species Therizinosaurus cheloniformis, combining the Greek therízo (scythe) and sauros (lizard) with chelóni (turtle), so the name translates to something close to "turtle-shaped scythe lizard." He was confident enough in the identification to erect an entirely new family, Therizinosauridae, just to hold it. The turtle interpretation was wrong in every particular except the claws' size, but the name has stuck for over seventy years.
It took 45 years to work out what kind of animal this even was
The turtle idea did not survive long. In 1970, Russian paleontologist Anatoly Rozhdestvensky compared the holotype claws to those of the carnosaur Chilantaisaurus and argued Therizinosaurus was a theropod dinosaur, not a reptile of any turtle-like kind; he also reidentified the metacarpal fragment as a metatarsal and reassigned the rib fragments to what he suspected was a sauropod. That same year, Polish paleontologist Halszka Osmólska and Ewa Roniewicz reached a similar theropod conclusion while describing the equally puzzling Deinocheirus from the same formation.
Confirming theropod status did not settle what family of theropod, and the classification kept drifting for two more decades. Rinchen Barsbold placed it among theropods again in 1976, and by 1979 to 1980 Altangerel Perle and Barsbold had grouped it with newly described relatives like Segnosaurus and Erlikosaurus into an infraorder called Segnosauria. Through the 1980s, though, paleontologist Gregory Paul argued the opposite case, that segnosaurs were not theropods at all but evolutionary intermediates between sauropodomorphs and ornithischians, a position other researchers including Jacques Gauthier and Paul Sereno backed for a time. The reversal came with two discoveries: the 1993 description of the far more complete Alxasaurus by Dale Russell and Dong Zhiming, which tied the segnosaur material firmly to Therizinosaurus and coined the superfamily Therizinosauroidea, and the 1999 description of Beipiaosaurus, a small, unambiguously feathered therizinosauroid from China, which placed the whole group among the feathered coelurosaurian theropods for good.
The claws, and the four-toed feet that go with them
Therizinosaurus is still known from only a handful of bones, but the ones that survive are extreme. Its manual unguals, the claw bones themselves, reach just over 52 centimetres in the fossil alone, making them the longest known from any land animal; a keratin claw sheath in life would have extended well beyond that bone length. The whole arm, from humerus through the radius to the tip of the second metacarpal, measured roughly 2.4 metres, with the humerus alone 76 centimetres long. Unlike the more strongly hooked claws of related therizinosaurs, Therizinosaurus's unguals were mostly straight and flattened side to side, curving only near the very tip.
Its feet were just as unusual for a theropod. Therizinosaurus stood on four functional, weight-bearing toes, a tetradactyl arrangement found in almost no other theropod lineage; the rest, Velociraptor among them, kept the ancestral three functional toes with the first reduced to a non-weight-bearing dewclaw held off the ground. The therizinosaur foot instead converges on the stockier, sauropodomorph-style foot of long-necked plant eaters, a body plan built for standing and browsing rather than running down prey.
What the claws were actually for: five decades of guesses, then a real test
Once the turtle idea was gone, the claws still needed an explanation, and researchers cycled through several before anyone tested one directly. Rozhdestvensky's 1970 paper suggested the claws might have torn open termite mounds or supported a fruit-eating diet. Barsbold in 1976 proposed they could have impaled prey or dug into loose ground, while noting how easily they seemed to fracture under load. In 1995, Lev Nessov floated a defensive role against predators, with juveniles possibly using the claws to climb trees the way modern sloths or hoatzin chicks do.
The first direct test came in 2014, when paleontologist Stephan Lautenschlager built digital models of several therizinosaur claw shapes and ran finite-element simulations, applying 400 newtons of force in three scenarios: scratch-digging, hook-and-pull, and piercing, published in Proceedings of the Royal Society B. Therizinosaurus's claws produced the highest stress of any shape tested, worst of all in the digging scenario, which argued against a burrowing or excavating function outright. The hook-and-pull scenario produced comparatively low stress, supporting a function centered on grasping and pulling vegetation within reach, though Lautenschlager could not rule out a secondary role in defense, intimidation, or gripping during mating.
A 2018 study by Scott Lee and Zachary Richards measured bending resistance in therizinosaur humeri and found them, like those of carnosaurs and tyrannosaurs, comparatively resistant to stress, which they read as support for forceful, robust arm use and floated a defensive claw function for slow-moving therizinosaurs that, unlike fleet-footed ornithomimosaurs, could not simply outrun a threat.
The most direct test came in 2023, when Bristol PhD student Zichuan Qin, working with Chun-Chi Liao of the Institute of Vertebrate Paleontology and Paleoanthropology in Beijing and Michael Benton, CT-scanned the claws, ran finite-element stress simulations, and compared the results against living animals with known claw functions, publishing in Communications Biology. The claws of Therizinosaurus came back with no identifiable mechanical function in any scenario tested. "These long, narrow claws were too weak for combat," Liao said in the university's press release. "Our engineering simulation shows that these claws could not withstand much stress... so we conclude that the largest claws of any animal ever were actually useless in mechanical function, and so must have evolved under sexual selection to be used in display. The adult Therizinosaurus I guess could wave the claws at a competitor and effectively say, 'look at me, back off,' or wave them around in some way like a peacock can use its tail in display to attract females for mating."
Jurassic World Dominion gave it a fight the fossils say it couldn't win
Therizinosaurus reached mainstream audiences through Jurassic World Dominion in 2022, where it is staged as a territorial predator that claws down a deer, corners Claire Dearing in a swamp, and joins forces with a Tyrannosaurus rex in the climax to kill the film's antagonist, Giganotosaurus, by driving its claws into it. It is the same kind of gap between blockbuster and fossil record that turned Dilophosaurus into a venom-spitting, frill-necked predator in the original Jurassic Park: a dramatic capability invented for the screen that the actual animal never had reason to use.
The Bristol team addressed the film directly in their 2023 press release: "We all saw Therizinosaurus in 'Jurassic World' hitting deer and killing the giant predator Giganotosaurus. However, this is unlikely," Liao said, pointing to the same finite-element results showing the claws could not withstand serious combat stress. In its actual ecosystem, Therizinosaurus shared the Nemegt Formation with the tyrannosaurid Tarbosaurus, one of the few predators large enough to threaten an adult of its size, and the evidence points to it avoiding that competition by browsing on vegetation out of easy reach rather than fighting anything for territory.
A body built for browsing, not for killing
The rest of the skeleton, reconstructed mostly from more complete relatives, points the same direction as the claw studies. Therizinosaurus carried a small, beaked skull (a rhamphotheca, the same horn-covered bill seen in birds and turtles, sat over the jaws) on a long neck, estimated at around 2.2 metres by researchers Mike Taylor and Matt Wedel using proportions drawn from the related Nanshiungosaurus. A wide torso housed a large gut suited to processing bulk plant matter, and sparse feathering, inferred through comparison with feathered relatives like Beipiaosaurus, likely covered the body. The pelvis was robust and angled backward in a way well suited to sitting rather than sprinting.
In 1993, Dale Russell and Donald Russell compared that body plan directly to Chalicotherium, an extinct herbivorous mammal with similarly powerful arms and a pelvis built for sitting, and to the feeding posture of modern gorillas, a case of unrelated animals converging on the same solution for reaching foliage. They proposed Therizinosaurus fed while seated, using its long neck and arms to pull branches within range without needing much force, and that its arms may have been long enough to brace against the ground and help push the animal back up from a resting position. A 2018 study by Anthony Fiorillo and colleagues, comparing jaw traits across therizinosaurids, further estimated a comparatively weak bite force in derived, later-branching members of the group like Therizinosaurus, consistent with cropping vegetation rather than the roughly 8,000-pound bite of a predator like Tyrannosaurus rex.