Before you freak out: you do not necessarily need a demo to pitch your game to a publisher. While it's true there's a lot that a demo can do to convey ideas and mechanics quickly and effectively, and of course every publisher may have different requirements for a pitch, you are not doomed if you don't yet have a demo ready for when you pitch.
That being said, however, if you're going to put in the time, effort, and resources in creating a demo to send to a publisher, there are a few things you can do (and ask yourself) to make sure all that work is showcased to its best advantage. Over the last few years I've played a lot of demos as part of game pitches, and some of them made a far better impression than others... for a wide variety of different reasons.
As always, this comes from my own experience in working for a publisher, and should not be taken as any sort of hard and fast rule set that every publisher will want you to follow. If the publisher you're speaking to gives you specific guidelines, be sure and follow those!
Put some meat on its bones.
Your demo doesn't need hours and hours of content, but it SHOULD have something resembling actual, functional gameplay. If what you're sending me is so bare bones it basically just qualifies as a technical prototype, you might want to wait and work on it more. What type of game are you making? Does the demo at least provide a taste of that expected gameplay? If you tell me you have the next greatest platformer but I don't get to do any actual, you know, great platforming, your demo might be too early. Don't rush it!
Provide information on what is a placeholder and what isn't.
It isn't always immediately indicative of what percentage of a demo's assets are final. You don't need to run down a list of things in exhaustive detail, but calling out the big ones can help make sure you're giving the potential publisher the right impression. This is good for you AND them, since you don't run the risk of them assuming your stiff placeholder animations represent the scope of your abilities, or that your simplistic sample puzzles are as challenging as the game is going to get.
Focus on what's important.
Identify your game's strongest pillars and shore those up. These key factors are what you need to get right above anything else for your game to succeed in its genre. If it's an RPG or adventure, your dialogue and characters need to be well-written and interesting. If it's a platformer or other action game, your movement should be smooth and responsive. Getting these elements right shows us that you understand what is most important to the foundation of your experience. By nailing these things, even if they might still require some polish or tweaking, even if everything else is temporarily animated with stick figures, chewing gum, and twine, you're focusing on the heart of your game, and that's what matters.
Respect their time.
Not every publisher or their team has the time to sit and play through your entire build, or even just sit through a lengthy opening tutorial or slow burn scene. Ask yourself this; "If I was showing this to someone who had never seen it before, and all I had was these fives minutes of gameplay to hook them into buying it... would they?" Offering a demo with debug options or the ability to hop around to key points can help you show off different mechanics or areas that wouldn't otherwise be seen, and allow them to get a better grasp of what the full experience of your project is like in a smaller amount of time.
Make sure it's playable.
You'll note that I said "playable" and not "bug free". You should, to the best of your ability, ensure that your demo runs as smoothly and as stable as possible. But you should also ensure that any mechanics and features are clearly explained, that any puzzle logic makes sense, that it's not horrendously unbalanced... you get the idea. We all expect and understand a certain amount of kinks and bugs in any build we get sent. But if we start running into big walls right off the bat, and they're things the developer should have spotted themselves, that doesn't give a great impression.
End it strong.
One of the best things you can do is leave us wanting more. Rather than giving us a demo that goes on forever (or until it gets too buggy/we get bored), give us a demo that has a stopping point that grabs us. An exciting boss battle. A big reveal. A challenging puzzle. Cap the experience when we're hooked and enjoying ourselves, and you stand a better chance of getting talks to proceed.