How to Write a Song: A Complete Guide for Any Skill Level

Most of the advice that exists on songwriting comes from the wrong place. They tell you where to find inspiration, grab your guitar, and let the music flow, as though every great song were born in a mystical flash of creative genius.

It's a good origin story. It's just not exactly how most songs actually come together.

This past year, 41,000 people wrote and produced over 123,000 songs on our platform. Not professional songwriters, mostly — just people with something to say. We expected love songs to dominate. We expected short, casual prompts. We were wrong about a lot of things. (More on that here if you want the full data.)

This guide is about that gap — between having an idea and having a song. That gap has narrowed, yet it still requires genuine choices about structure, words, and melody — whether you're starting on a guitar or opening an app on your phone.

Write About What Hurts

Across those 123,000 songs, the ones about depression and anxiety got the most engagement. Not love songs. Not party anthems. Songs about pain — played 1.74 times on average compared to 1.15 for birthday songs and 1.48 for love songs.

Breakup songs came second. Then wedding songs. Then, oddly, songs about grief.

I'm not quite sure what to make of this, to be honest. Part of it is probably selection bias; someone processing a real feeling is much more likely to replay what they've written than someone making a fun birthday jingle. But it's what the academic research has been telling us for years now. In 2024, a paper in Scientific Reports analyzed 12,000 songs in the English language over five decades and found that lyrics have become steadily more personal and more emotionally raw — and that listeners respond to the shift. Another study from the University of California, Irvine, this time on 500,000 songs, found that while music has gotten sadder since the 1980s, the tunes that actually break through are more emotionally intense than average, in either direction.

The point isn't that you should write sad songs. It's that whatever you're kind of scared to say out loud is probably the thing worth writing about. A song about sitting in your car in the driveway because you're not ready to go inside yet will always beat a song about how love is beautiful.

Most likely, you already know what your song is about. Start there.

Structure Isn't Optional — It's the Whole Game

It's probably the least sexy part of songwriting. Probably the most important, too.

Songs in our dataset without an identifiable verse-chorus structure averaged 0.19 plays. Structured songs averaged 1.3 to 1.6 plays. That's not a quality difference. That's the difference between what registers as a song and what doesn't.

One could argue this is partly attributable to the nature of AI-generated music. And maybe it is, to some extent. But researchers at Queen Mary University of London looked at 17,000 Billboard Hot 100 songs over fifty years and found something strikingly similar: while production styles went through three complete revolutions — British Invasion in '64, synth-pop in '83, hip-hop in '91 — the structural formulas underneath changed hardly at all. Verse-chorus-verse-chorus-bridge-chorus has been the dominant skeleton for over sixty years.

Each section has a specific job:

• Verse sets the scene. Characters, setting, situation.
• Chorus delivers the emotional center — what you want someone to feel.
• Bridge breaks the pattern. Shifts perspective. Says what you've been holding back.
• Pre-chorus builds tension so the chorus lands harder.

What did change: songs are getting to the point faster. Hit Songs Deconstructed found that chart-toppers opening with the chorus before any verse jumped from 25% to over 40% in recent years. If your hook is your strongest asset, you might not need a two-verse warmup to get there.

Pick your structure before you write a word. Not because it limits creativity — because it focuses it. Filling in rooms is easier when you've already drawn the floor plan.

The Chorus Problem

The chorus is the song. Everything else exists to set it up or bring it back.

A study from USC and Bocconi University tracked every Billboard Hot 100 entry between 1958 and 2012 and found that more repetitive choruses correlated with higher chart positions. The explanation is "processing fluency" — the brain generates a positive response when it grasps something easily. We enjoy what we can predict.

I think this gets oversimplified into "make it catchy." A better way to think about it: your chorus should feel inevitable. Like it's the only thing you could have said after the verse set it up. Take your central idea and compress it to 2-4 lines. Strip away everything that isn't the core feeling.

The test I keep coming back to: can someone sing it after hearing it twice? Not perfectly — just the shape of it. If not, it probably needs to be simpler.

Lyrics: Be Specific or Be Forgotten

"You broke my heart" has been written a million times. It means nothing anymore.

"I still check your Spotify to see what you're listening to" — that means something.

The difference is specificity. Concrete imagery. Details that feel true even if they're invented. A study of song concreteness over time found that the eras where popular lyrics used more concrete language — specific objects, places, sensory details — corresponded with periods of higher listener engagement.

We saw something like this in our own data, though I want to be careful about reading too much into it. When we compared prompts behind zero-play songs versus songs that got 10+ plays, there was a clear pattern:

The high-play prompts used words like "time," "back," "said," "maybe," "chorus," "verse." They read like rough drafts of actual songs. People telling stories.

The zero-play prompts used words like "should," "want," "style," "beat," "name." They read like orders placed at a counter.

That might just mean invested writers produce better songs — which, sure, of course. But I think it's also about the writing itself. When you're arranging real words in a real sequence with real feeling, the result is different from when you're describing what you want from a distance. The former is songwriting. The latter is a brief.

Writing the Verses

Your first verse drops the listener into the world. Where are we? Who's talking? What just happened?

Your second verse deepens it. Don't restate the first — advance. Reveal something. Change the angle.

Common traps:

Forced rhymes. "Home" and "alone" work. Don't contort your meaning to land on "phone." A near-rhyme that says what you mean beats a perfect rhyme that says nothing.

Overwriting. That 2024 Scientific Reports study found vocabulary in popular songs has gotten simpler across every genre for fifty years — and simpler songs perform better. This isn't dumbing down. It's editing. One sharp image beats three mediocre lines.

If lyrics are your sticking point, we have a deeper guide on how to come up with song lyrics.

Melody Is the Part Nobody Can Teach You (But Here's What We Know)

This is the hardest section to write honestly, because melody is where craft starts to blur into something less explainable.

What science can tell us: neuroscientists at McGill University used PET scans to show that music triggers dopamine in the brain's reward system — the same circuitry as food and sex. The interesting part is when. Not at the musical climax. During the anticipation of it. The build, the tension before the chorus drops — that's where the brain lights up most.

A Durham University study surveyed 3,000 people about earworms and found that melodies which get stuck in your head tend to follow common pitch contours — rising then falling — but feature one or two unusual intervals at their turning points. Familiar enough to process. Surprising enough to snag.

What science can't tell you: why one melody over the same four chords as "Let It Be" and "Someone Like You" makes you feel absolutely nothing and another one makes you pull over your car. I don't think anyone entirely gets that. But you can put yourself in the way of finding out.

With an Instrument

Loop 3-4 chords. Sing over them. Nonsense syllables, placeholder words, whatever surfaces. Don't judge it — just record everything on your phone, because your first pass is probably closer to right than your tenth.

Hooktheory analyzed 1,300+ popular songs and found that four chords — C, G, Am, F (or their equivalents in any key) — account for a huge proportion of hits. A deeper study of 464,000 recordings confirmed that chord diversity in popular music has been narrowing since the 1960s. Not because songwriters got lazier, but because the progressions that resonate keep resonating.

• I – V – vi – IV (C – G – Am – F) — the pop workhorse
• vi – IV – I – V (Am – F – C – G) — same chords, darker starting point
• I – IV – V – I (C – F – G – C) — rock and country bedrock

Without an Instrument

This was where songwriting stopped for many people, for a long time. You'd have words but no way to hear them as music. The only option was to find a collaborator, book studio time, or spend months learning guitar — and by then the impulse that made you want to write the song might be long gone.

That's changed. AI song generators let you paste in lyrics, pick a genre and mood, and hear a produced track in minutes. It works best with vocal-driven genres like acoustic, rap, country, and indie — and it keeps getting better across the board.

The most surprising thing we found looking at our own data: 96.8% of the songs created were originals with user-written lyrics. The average prompt was 146 words. Over a third of users wrote prompts exceeding 1,000 characters — with full lyrics, verse/chorus markers, and production direction. These aren't people asking a machine to write for them. They're writers using a new kind of instrument.

Both approaches coexist. Plenty of people sketch ideas on guitar, then hear their lyrics as a country arrangement or a rap track they couldn't have produced themselves. The words don't change. The possibilities do.

Genre Changes the Rules More Than You'd Think

The fundamentals apply everywhere. But each genre has its own gravity, and ignoring it usually means the song doesn't quite work — even if you can't exactly explain why.

Rap is rhythm before everything. The best rappers don't start with words. They tap out a cadence, fill it with nonsense syllables, then swap in real lyrics once the rhythm feels right. Internal rhyme matters more than end rhyme. "I stack my words and pack them tight, every track I write hits right" — four rhymes in two lines, none at the end. That's the craft. You can feel when someone's done this work and when they haven't.

Country thrives on storytelling, which sounds obvious until you realize how few genres actually require it. The best country songs are basically short films — grounded in a specific place, told in language so conversational it sounds overheard. What separates great from good is, nine times out of ten, the turn: a line in the bridge or final chorus that reframes everything before it. (This is also, incidentally, the genre that did best in our 123K-song dataset. Country tracks averaged nearly double the plays of the overall average, and country pop had a 95% like rate. I suspect it's because storytelling demands specificity, and specificity is what makes songs stick.)

Love songs come down to scope. Pick one moment, not the whole relationship. "You're my everything" evaporates on contact. "You squeeze my hand three times in the dark" is something only you would know, and that's exactly what makes it work.

Pop is economy. If the chorus doesn't stick on the first listen, it probably needs to be simpler, not cleverer. Pop melodies tend to move in small steps with one unexpected leap for emphasis. There's a reason the biggest pop songs sound "obvious" — they're optimized for the brain to grab onto instantly.

Editing Is Where the Song Actually Happens

Most people think writing is the hard part. It's not. Editing is.

Give it at least a day. What felt profound at midnight often seems obvious by morning. Sing everything out loud — lyrics that read well on paper have a way of tripping you up mid-breath when you actually try to sing them.

The rule I find hardest to follow: one idea per song. If you've got three ideas fighting for space, that's three songs, not one. Today the average Billboard hit has 4.5 credited writers, and even with all those people in a room, the goal is still compression. Not more ideas. Fewer, better ones.

One thing that's underrated: play the song for exactly one person and watch their face. Not a crowd. Not a text with a voice memo. One person in the room. You'll know which parts work before they open their mouth.

Getting It Out of Your Head

If it only exists in your head, it's not a song yet.

Record it. Voice memo, GarageBand, BandLab, an AI tool that turns lyrics into a produced track — the method matters less than the fact that it now exists outside your skull. You can share it. You can hear what doesn't work. You can build on it tomorrow instead of trying to remember it.

Professional songwriters churn out 100 to 150 demos a year. Most go nowhere. The ones who get good aren't the ones who write one perfect song. They're the ones who finished enough bad ones to learn what a good one feels like.

Here's a strange thing we noticed in our data: the most-played song across all 123,000 was a hip-hop diss track roasting the Gen Alpha "6-7" meme. Meanwhile, plenty of technically polished songs sat at zero plays. I'm not entirely sure what the lesson is — maybe that the best songs come from an impulse so specific and immediate that you can't not make them. Or maybe that diss tracks are just inherently funny. (They had an 85% like rate across the whole dataset. People enjoy being roasted.)

Either way: don't overthink it. Write the song. Record it badly if you have to. The one you finish teaches you more than the one you're still planning.