The answer depends on the type of utilitarianism involved. It is possible for cheating to be mandatory for an act utilitarian.
For act utilitarianism an agent is required to do the action that maximises the general welfare, where the general welfare is construed in terms of aggregate or average welfare. Or on an alternative approach, an act utilitarianism requires an agent - any agent, all agents - to do whatever action, in a particular situation for action, maximises the benefit of the greatest number.
The key is that act utilitarianism takes feasible actions one by one, and regardless of their nature or description (cheating, returning a favour, breaking a contract, supporting a charity ... whatever) takes moral account purely of their(probable) consequences, and requires the agent to do that action which produces (or is most likely to produce) maximally beneficial consequences. There is nothing in the logical form of act utilitarianism, concentrating as it does solely on the consequences of a particular action, that prevents cheating from being the action which produces (or is most likely to produce) maximally beneficial consequences in a particular situation for action. Depending on its consequences, which alone are relevant for act utilitarianism, cheating could be mandatory. Could be - and in a different situation for action, with different (probable) consequences, cheating could be wrong. There is (by definition) no generalising for act utilitarianism.
In contrast, rule utilitarianism requires an agent to act on that rule of action which, if generally acted on, produces (or is most likely to produce) maximally beneficial consequences. Alternatively, maximises the benefit of the greatest number. A 'rule of action' here identifies a particular type of action, e.g. cheating, lying, promise-breaking, giving to charity, returning a favour. There is extensive agreement among rule utilitarians that cheating is not such a type of action, if generally acted on, and so rule utilitarianism normally proscribes lying.
Note: the different formulas ('if generally acted on, produces (or is most likely to produce) maximally beneficial consequences', or 'maximises the benefit of the greatest number', and others besides) are not the key point. The key point is whether it is individual actions (act utilitarianism) or types of action (rule utilitarianism) that are morally mandatory.
Harry J. Gensler, Ethics, London: Routledge, 1998: 150-2.
Bernard Rosen, Ethical Theory, London: Mayfield, 1993: 99-101.
Geoffrey Thomas, An Introduction to Ethics, London: Duckworth, 1997: 72-4.