So, the relationship of speeches in the ancient historians to “what really happened” is very difficult to understand, and while it’s commonly taught to undergrads (including myself) that the historians “made up” the speeches, it’s a very unfair characterization.
The classic text about method of treating speeches is Thucydides, 1.22:
Insofar as these facts [his account of the Peloponnesian War] involve what the various participants said both before and during the actual conflict, recalling the exact words was difficult for me regarding speeches I heard myself and for my informants about speeches made elsewhere; in the way I thought each would have said what was especially required in the given situation, I have stated accordingly, with the closest possible fidelity on my part to the overall sense of what was actually said.
This passage is often taken to mean that all Thucydides’ speeches are made up; imaginary speeches that suit the characters and the situation, like in a play. But like almost everything Thucydides says, someone else thinks it means the exact opposite. Donald Kagan (although there is a long bibliography on what this phrase means – and whether Thucydides sticks to his stated method) most publicly argued that this means Thucydides did not make up his speeches – he presented his speeches with the greatest possible fidelity to what what actually said, as far as he or his witnesses could remember! This obviously is problematic: sure, the Funeral Oration might be well-remembered with many witnesses to give accounts, but something like the Melian dialogue? How does he know what was said in a private conference in a far-away place? To many it seems shocking that Athenian ambassadors would be as brazenly cynical while trying to win an ally as Thucydides presents them. I happen to like Kagan on this point; I think speeches are as good as the author’s evidence for them.
So what evidence for actual speeches might have been available? Thucydides has already mentioned memory and witnesses – a limited help, but some help at least. Orators, both in Athens and in Rome, published speeches. Demosthenes and Cicero are the most famous, but there are others that survive. Speeches in Rome at least (an example in Greece doesn’t come immediately to mind) could also be inscribed. Pliny the Elder, Natural History, VII.28.104-106, gives a biography of the Roman war hero M. Sergius Silus, who fought in Punic II. He attributes his knowledge of Sergius’ life to a speech Sergius gave defending himself from ejection from a priesthood, which had been inscribed and survived to Pliny’s day.
Tacitus took advantage of at least one such inscription when he wrote his account of Claudius’ speech to the Senate in 48 CE. An apparently verbatim inscription appears in the Lyon Tablet. Tacitus gives a version of this speech at Annales 11.23-24. It’s clear to me that Tacitus is familiar with the text as represented in the Lyon Tablet (though some scholars are more interested in how Tacitus departs from the Tablet), at least as far as the arguments go. He does, however, restructure the argument, cut out superfluous bits, and does general tightening. I’m just going to produce two passages as a demonstration of the sort of thing Tacitus is up to.
In Claudius’ speech as given in the Lyon Tablet, Claudius presents two objections to his motion that Gauls should be admitted to the Senate:
But, you may say, is not an Italian senator more useful than a provincial one?
and
It is time now, Tiberius Caesar Germanicus, to reveal to the senators where your speech is headed; for you have already come to the extreme limits of Gallia Narbonensis.
This last comment is a concern about the geographical reach of the new Senators Claudius proposes. Tacitus took these objections out of the speech proper and puts them in the mouths of Claudius’ opposition:
In the consulship of Aulus Vitellius and Lucius Vipstanus the question of filling up the Senate was discussed, and the chief men of Gallia Comata, as it was called, who had long possessed the rights of allies and of Roman citizens, sought the privilege of obtaining public offices at Rome. There was much talk of every kind on the subject, and it was argued before the emperor with vehement opposition. "Italy," it was asserted, "is not so feeble as to be unable to furnish its own capital with a senate. Once our native-born citizens sufficed for peoples of our own kin, and we are by no means dissatisfied with the Rome of the past. To this day we cite examples, which under our old customs the Roman character exhibited as to valour and renown. Is it a small thing that Veneti and Insubres have already burst into the Senate-house, unless a mob of foreigners, a troop of captives, so to say, is now forced upon us? What distinctions will be left for the remnants of our noble houses, or for any impoverished senators from Latium? Every place will be crowded with these millionaires, whose ancestors of the second and third generations at the head of hostile tribes destroyed our armies with fire and sword, and actually besieged the divine Julius at Alesia. These are recent memories. What if there were to rise up the remembrance of those who fell in Rome's citadel and at her altar by the hands of these same barbarians! Let them enjoy indeed the title of citizens, but let them not vulgarise the distinctions of the Senate and the honours of office."
It deals with the same objections, but removes them from the middle of the speech (where they are in the Lyon Tablet) and Tacitus makes Claudius’ speech respond to them, making both the objections and arguments more forceful.
This is an interesting move by Tacitus, because the general thought is ancient historians did not produce speeches that were elsewhere published. It may suggest that, while Tacitus was aware of the content of the speech, the inscription was difficult to use or not well known, and so he felt fine reproducing an edited version. Sallust took what we think of as the usual approach: in his Catilinarian Conspiracy, he didn’t give anything about Cicero’s four great speeches but passed over them in favor of the speeches of Caesar and Cato (and a speech delivered in secret by Catiline – and again, we can wonder what his source for that was). Cicero’s speeches of course were widely published by Cicero himself. Sallust did not feel comfortable reproducing them or versions of them in his own work.
Another source in the Roman period was the Acta, the daily record of public events. These were strange documents we don’t understand terribly well, but they seemed to be a sort of daily political bulletin reporting court cases and other public business. They don’t seem to be used very often as a source, but Asconius used them in his commentary on Cicero’s speeches, so some people were looking at them. One thing we learn from Asconius’ commentary on Cicero’s Pro Milone (37C) is:
Q. Pompeius, C. Sallustius [That’s Sallust, our non-plagiarizing historian!], and T. Munatius Plancus, tribunes of the plebs, were among the first to hold contiones that were extremely hostile toward Milo, and calculated also to arouse animosity against Cicero for his strenuous efforts to defend Milo.
We imagine this sort of detail was taken out of the Acta. Contiones were public meetings where magistrates addressed the People of Rome. Here we have the speakers and a summary of the content.
So ancient historians had all of these things to help them not make up speeches, but of course the verisimilitude of the speech can only be as good as the sources they have to work with. Livy’s early books, for instance, have speeches it’s very hard to believe are grounded in any meaningful evidence. These situations provide an opportunity for a more imaginative piece, suited to the needs of the moment and the character of the speakers, as the careless reader of Thucydides might think he means to say is his method. There are some scholars who think Livy is pulling those sorts of speeches from actual drama, or if they are original he’s imitating drama in those moments.
As I said, there is a huge bibliography on this question, with scholars thinking the speeches are more or less real, as well as arguing about specific speeches. The important thing, I think, is when confronted with a speech in an ancient historian, to consider how the historian might know what they report, and to build your ideas about whether the speech was made up from there.