In his sprawling new book Tom Holland undertakes to explain nothing less than the origin of Islam. This is a subject as relevant to today's world as it is controversial within it. How Islam began was obscure right from the start, above all to the surprised Christians who first succumbed to the Arab armies that surged out of the Arabian peninsula in the seventh century. They had seen themselves as confronting a different threat. After all, the Persians had captured Jerusalem in 614 and soon moved into Egypt. At that moment they appeared to be the principal antagonist of the Byzantine empire based in Constantinople. No one could have imagined that a little over two decades later the Persian empire would be in its death throes and that the Patriarch of Jerusalem would be turning over the city to an Arab caliph.

The beginnings of Islam have always been anchored in Mecca in the northwestern part of the Arabian peninsula. Here Muhammad was believed to have received from the angel Gabriel the earliest revelations that became incorporated in the Muslim scripture, the Qur'an. Scholarly debate about the revelations and about Meccan society has gone on for centuries, but no one before has seriously doubted the conjunction of Muhammad and Mecca. Holland wants us to believe that Muhammad did not come from Mecca at all but from southern Transjordan, and that his revelation was a compound of languages and ideas floating around in the Near East.

Holland came to his work on Islam unencumbered by any prior acquaintance with its fundamental texts or the scholarly literature. He modestly compares himself to Edward Gibbon, whom he can call without the slightest fear of contradiction "an infinitely greater historian than myself". In the Decline and Fall, at the opening of his magisterial chapter 50 on Muhammad, Gibbon had candidly acknowledged his ignorance of "Oriental tongues", but he also expressed his gratitude "to the learned interpreters who have transfused their science in the Latin, French, and English languages". Holland seems to have confined himself largely to interpreters, learned or otherwise, writing in English, but his efforts to inform himself, arduous as they may have been, were manifestly insufficient.

He has written his book in a swashbuckling style that aims more to unsettle his readers than to instruct them. I have not seen a book about Arabia that is so irresponsible and unreliable since Kamal Salibi's The Bible Came from Arabia (1985). Although that work was depressingly misguided in replacing biblical places with their homonyms in the Arabian peninsula, it at least revealed an accomplished scholar who had gone badly astray. Holland has read widely, but carelessly. He starts out with an irrelevant, though arresting, account of a defeated Jewish king in Arabian Himyar (Yemen) killing himself by riding his horse into the Red Sea. It is typical of Holland's style to lead off with this fanciful story when an inscription from the time of the king's death records that the Ethiopians killed him.

Holland explodes with indignation over the traditional term, jahiliyya (age of ignorance), for the time before Muhammad. After a tabloid view of Arab culture in that period, he declares: "The effect of this presumption was to prove incalculable. To this day, even in the west, it continues to inform the way in which the history of the Middle East is interpreted and understood." This was partially true in Gibbon's time, but it is quite false today. Research and publication on pre-Islamic history, archaeology, art and languages may be found in many western universities, such as Oxford, as well as in many Arab countries, including Saudi Arabia, Jordan, Lebanon, Egypt, and Syria.

The past 30 years have seen lively controversies in the scholarship on early Islam, much of it emanating from the revisionist work of John Wansbrough in analysing the text of the Qur'an and its possible links with both Christian and Jewish language and thought. This is catnip for Holland, as is the revisionist work by Wansbrough's disciple, Andrew Rippin, and, much more idiosyncratically, by the pseudonymous Christoph Luxenberg, who dares not speak his name. Although these debates are all solidly grounded in close textual study, they can do little more than titillate uninitiated readers because the dust has not yet settled.

Holland's failure to follow Gibbon in examining French scholarship means that he has missed many of the most important recent discoveries, above all the large number of inscriptions from late antique south Arabia that Christian Julien Robin and his associates in Paris have been publishing in a steady stream. We now know much more about the Judaism of Himyar, the conflict with Christian Ethiopia and the Persian occupation of western Arabia. In discussing early Qur'an manuscripts Holland has missed the collaborative manuscript, in five different hands, which François Déroche has dated to the third quarter of the seventh century. It appears to antedate the Qur'anic inscriptions in the Dome of the Rock in Jerusalem.

The scattershot nature of Holland's investigations is particularly apparent in his breezy reference to the Qur'an manuscripts that were found in Sana'a, Yemen, in 1973. He hints darkly at censorship to explain publication delays caused by textual variants in a palimpsest but is unaware that the palimpsest itself and two other manuscripts are actually now with the publisher. He is also unaware that a second cache of Qur'an manuscripts was discovered five years ago in renovations of the Great Mosque in Sana'a and that in February 2010 the Yemeni authorities granted permission for them to be studied.

But Holland is at his most irresponsible when he turns to the Meccan origins of Islam. After reasonably supporting Patricia Crone's argument against the tradition of Mecca as a mercantile centre, he goes on to ask whether the place itself might not be an invention in the story of Muhammad. He raises the possibility that the Qur'anic pagans, called mushrikun, might be confederate tribes simply because the word is constructed from the Arabic root for "sharing". He looks for these tribes in southern Jordan and not only thinks of placing Muhammad among them but proposes that his own Meccan tribe, the Quraysh, took its name from the Syriac word qarisha, which, according to Holland, would have been "duly Arabised". This jaw-dropping idea depends on Holland's mistaken view that the Syriac word could allude to a confederation. What it means is to clot or congeal.

For some reason Holland's book was released in the Netherlands in Dutch before it appeared in English. It had a different title then, The Fourth Beast. A marketing strategy of this kind looks like a conscious effort to profit from recent Dutch anxiety over Muslim immigrants. But Holland's cavalier treatment of his sources, ignorance of current research and lack of linguistic and historical acumen serve to undermine his provocative narrative. In the Shadow of the Sword seems like an attempt by author, agent and publisher to create a very different account of early Islam, but fortunately the quality of the book stands in the way.

 Glen Bowersock's From Gibbon to Auden: Essays on the Classical Tradition is published by Oxford.

In 706AD the caliph al-Walid decided to commission a building as a centrepiece for his new capital, Damascus. Only 74 years had passed since the death of the prophet Muhammad; the Arabs' new empire was still in the making, and there was no such thing as imperial Islamic architecture. The caliph found inspiration for his mosque in both Christian and pagan temple architecture. And while he built it on the site of one of the greatest of all Roman temples, demolishing the Christian church that stood inside the precincts, he incorporated many of the Roman stones, as well as the tomb of John the Baptist. For decoration, he brought Byzantine craftsmen over to piece together the vast gold mosaics.

The idea of borrowing is untroubling in architecture  we expect to see continuity and evolution in buildings. But the idea that religions evolve out of one another is more disturbing. Christians have choked on the notion that many of their rituals were borrowed from pagan rites. And heaven help the historian who dares to suggest that Islam might be a product of earlier religions and not, as the faithful insist, a revelation direct from God. Tom Holland has done exactly this in his brilliantly provocative new book  and we must hope that heaven is smiling on him now.

Holland is the author of two extremely successful works of ancient history  successful in both the creative and the commercial senses. In Rubicon, he traced the end of the Roman Empire. In Persian Fire, he focused on the fifth-century BC conflict between the Persians and Greeks, between east and west. In the Shadow of the Sword is a more ambitious and more important book.

The Roman and Persian empires are traditionally seen as collapsing into a void that our schoolbooks called the Dark Ages. Holland's thesis is that there was no void: depending on how you interpret the material, the decline of those empires led to the rise of the Arabs, or the rise of the Arabs led to their decline. To test this idea, Holland ranges around the centuries. Not content with bringing us the end of Rome, he mentions its foundation as well, something he also does with Judaism and the Israelites, with Jesus and Christians, with Zoroaster and the Persians. He is a restless wanderer across the ancient world, both geographically and intellectually. Abraham, Isaac and Moses have walk-on parts, as do the emperors Constantine and Valerian, the empress Helen, Saint Simeon the Stylite (both of them, for it appears there were two) and a dazzling range of other characters from late antiquity. There are so many of them  and few stay around long enough for a reader without prior knowledge to get a real grip on them  that the narrative risks being overwhelmed by the comings and goings. But Holland is a skilful and energetic narrator, and while he guides us along the more intricate twists and turns of the period, he also keeps our eyes on the bigger story  on the revolution that brought down the old order and ushered in the new. For this was the period, Holland says, when allegiance to land, tribe or state was superseded by devotion to one almighty deity.

The assertion that Rome and Persia declined because of "the revelation of the word of God to His Prophet in far-off Mecca" is likely to be less controversial than Holland's examination of the details of the life of that prophet. He counters the widely accepted  and, to most Muslims, inviolable  view of Muhammad's life and the revelation of the Qur'an with what he calls "the traditions of secular scholarship". These traditions are based on an insistence on hard evidence and a demand that such evidence be scrutinised. The problem with the life of Muhammad is that there is almost no textual support for it until almost two centuries after his death. Holland is not going so far as to say that he never existed. Just that the account we have comes from the 800s, by which time Haroun al-Rashid was caliph over an empire that stretched from China to the Atlantic, and the 1,001 Nights was being compiled. The veracity of details and perhaps also some of the more important moments of the prophet's life were  are  impossible to prove.

The same goes for the Qur'an. There is no written mention of it in the period immediately following Muhammad  nor any commentary on it until the eighth century. Holland does not, perhaps, give enough weight to the value of oral tradition, the route by which the Qur'an is said to have been received (Muhammad is widely held to have been illiterate) and initially preserved. But his investigation does turn up some exciting possibilities as to the origin of the text and also to the location of Mecca: Holland points out that "there is not a shred of backing" in the Qur'an for locating Mecca in the Hijaz. The first text-based claim for its location in what we now know as Mecca, written a century or so after the revelation of the Qur'an, was a preposition "taken wholly for granted". He might be right, he might be wrong, but there is no denying that the challenge is both stimulating and, in a world of increasingly rigid Muslim dogma, refreshing.

The Qur'an anticipated the day of Holland's coming (or someone very like him). Sura 25 instructs Muslims to counter the claim that "these are fables of the ancients which he has got someone to write down for him" with the insistence that it was "revealed by Him Who knows every secret". For believers, these words are proof enough of the veracity of the Qur'an. Some have gone further and used them as justification for intellectual, legal and physical attacks on people who claim otherwise. The lives of some people who have dared to question the historicity of the prophet Muhammad and the Qur'an have been ruined, even ended. We must hope that Holland is spared their wrath and that his excellent book will be lauded, as it should be, for doing what the best sort of books can do  examining holy cows.

Anthony Sattin's A Winter on the Nile and Lifting the Veil are both out in paperback