Stop Agonising Over Website Copy and Just Ship the Thing

The old way took me three months

There was a time when getting a 12-page site live would take me three months.

Not three months of coding. Not three months of design. Three months of agonising over copy.

I would sit there trying to make every line perfect. I wanted it to flow. I wanted it to sound intelligent without sounding forced. I wanted it to feel like a real reflection of me and my work. I wanted every section to say exactly the right thing in exactly the right way. I wanted the whole site to feel coherent, persuasive, and polished enough that, if the right person landed on it, they would instantly “get it.”

And, to be fair, that instinct is not stupid. It comes from a good place. Positioning matters. Clarity matters. Language matters. A confused website is expensive. Bad copy creates drag. Weak framing attracts the wrong people and repels the right ones.

But there is a point where care turns into avoidance.

That was the trap for me.

Because, if I am honest, a lot of what I called “craft” was actually delay with a sophisticated excuse attached to it. I was not just trying to write well. I was trying to remove uncertainty before launch. I was trying to solve conversion in theory. I was trying to pre-approve the market’s reaction before the market had even seen the page.

That is impossible.

You cannot solve conversion in theory before the market has seen the page.

What changed

Now the process looks very different.

Now I can dump a bunch of docs into Claude, see what it spins up, realize it is roughly as good as what I probably would have produced after weeks of overthinking, get the project live, and iterate from there.

Is it perfect? No. Does it misunderstand context sometimes? Yes. Does it need editing? Obviously. But it is live. That changes everything.

Because once the page is live, I stop dealing in imagination and start dealing in feedback. I can look at scroll depth. I can look at bounce rate. I can see whether people click the CTA. I can see whether a section is getting ignored. I can watch what happens when I change a headline. I can compare one framing against another. I can stop treating copy as a private act of self-expression and start treating it as part of a commercial system.

That is a much healthier way to work.

The core problem: most copy decisions are made too early

The thing people do not like admitting is this: you do not really know what copy will convert until it is in the wild.

You can have good instincts. You can know your product deeply. You can have years of experience. You can understand brand, positioning, direct response, UX, design, sales conversations, customer pain points, market sophistication, and all the usual frameworks, but you still do not know.

Because there is a difference between a plausible theory of resonance and actual resonance. There is a difference between what sounds right in your head and what causes a stranger to keep reading. There is a difference between what you think is obvious and what is actually legible to someone with none of your internal context. That is where a lot of website copy goes wrong.

The founder or operator is standing too close to the thing. They know what the product does. They know what is unique. They know why the offer matters. They know what the words are pointing at. So they write a headline and a short paragraph and think, “Well obviously that communicates the value.” But it often does not.

Not because the prospect is stupid. Because the prospect is busy, distracted, skeptical, and reading with none of your background knowledge.

Crucially, they are usually not reading in the way you imagine they are reading. They are scanning. They are looking at the largest text first. They are pattern-matching against other sites they have seen.

They are trying to work out, very quickly:

• Is this for me?
• Is this credible?
• Is this specific?
• Is this expensive?
• Is this worth my time?
• Do I understand what this person actually does?

That is the real environment your copy is operating in. Not a perfect reading environment. Not an attentive seminar room. Not a charitable audience carefully parsing your nuance. A skim-heavy, context-poor, high-distraction environment.

Once you accept that, a lot of “perfect copy” turns out to be badly allocated effort.

The market is not reading your site the way you read your site.

Why AI matters here, and why the hype still misses the point

I do not think the value of AI is that it writes perfect copy.

It does not. I do not think the value is that it deeply understands your market by default. Usually it does not. And I definitely do not think the value is that human judgment no longer matters. If anything, human judgment matters more.

What AI changes is the cost and speed of first-draft production.

That matters because first drafts are not where most commercial value comes from anyway. In most cases, the value comes from the loop:

1. Get something viable live.
2. Observe how real people respond.
3. Edit based on signal.
4. Repeat.

AI compresses step one. That is the real win.

It gets you to a draft with structure, language, and momentum. It gives you raw material. It reduces the activation energy required to turn scattered thoughts, notes, decks, sales docs, and offers into a page that can actually exist on the internet and start generating evidence.

That is much more important than whether every sentence is beautiful. Because beautiful but unpublished copy has no commercial function. Published, imperfect copy can still teach you something, and that learning compounds.

The technical reality: a website is not copy, it is a system

One of the biggest mindset shifts for me has been realizing that copy is not a standalone artifact. It is one layer inside a broader conversion system.

A page is not just words. It is:

• message hierarchy
• information architecture
• design emphasis
• visual pacing
• CTA placement
• trust signals
• mobile readability
• loading speed
• traffic intent
• audience temperature
• offer clarity
• the sequence of assumptions the user has to resolve before taking action

That matters because people often blame “copy” for problems that are really structural.

Sometimes the headline is weak. Sure. But sometimes the real issue is that the page asks the visitor to process too many ideas too early.

Sometimes the CTA is too abstract. Sometimes the proof is buried too low. Sometimes the offer is unclear. Sometimes the page is trying to speak to three audiences at once. Sometimes the design makes the important part look visually unimportant. Sometimes the page is mobile-hostile, so no one even gets far enough to experience the copy properly. Sometimes the copy is perfect, the audience targeted and the page beautiful, it was just the expectation that was wrong.

This is why over-polishing wording in isolation can be such a waste. You can spend six hours on a paragraph that is sitting inside a page with the wrong hierarchy, wrong sequence, wrong emphasis, and wrong audience assumptions. No amount of sentence-level genius fixes system-level confusion.

That is also why AI can be more useful than people think, even when its prose is not brilliant. It can help you externalize the system faster. It can force the rough structure into existence. Then you can inspect the actual object instead of mentally simulating it.

That is a much better working surface.

Sometimes the copy is perfect, the audience targeted and the page beautiful, it was just the expectation that was wrong.

Why this is emotionally hard if your work is close to your identity

There is another layer to this, and I think it matters more than people admit.

If your business is close to your identity, website copy does not feel like ordinary writing. It feels personal.

You are not just describing a service. You are describing yourself, your value, your way of thinking, your taste, your standards, your competence, maybe even your future. So every sentence starts carrying psychological weight far beyond its literal role on the page.

That makes editing hard.

Because once the copy becomes identity-adjacent, bad process creeps in:

• You defend lines because they feel true to you, not because they are useful to the reader.
• You over-explain because you want to be fully understood.
• You keep nuance that should be cut because it feels important.
• You treat simplification like misrepresentation.
• You conflate personal precision with market clarity.

I know this tendency very well.

That is one reason AI helps, even when the output is imperfect. It introduces emotional distance. It gives you something to react to rather than something to protect.

That is incredibly useful.

Editing a flawed draft is easier than trying to summon the perfect expression of yourself from a blank page. When the text is not born directly from your nervous system, you can be more ruthless. You can cut faster. You can see the structural issues more clearly. You can ask better questions. Not, “Is this me?” But, “Does this work?” That is a more commercially useful question.

You can ask better questions. Not, “Is this me?” But, “Does this work?”

What actually happens after you publish

This is the part that changed my thinking the most. Before a page is live, every decision feels final. After it is live, you realize almost nothing is final.

The headline can change. The section order can change. The CTA can change. The imagery can change. The proof can change. The offer framing can change. The page length can change. The positioning can change. And often, it should.

Publishing is not the final exam. It is the start of measurement. Once you internalize that, the goal shifts from “make this perfect before launch” to “make this legible enough to generate signal.” That is a much better signal because it tells you whether your assumptions were right, but usually they were not.

What you thought was the most important feature might barely matter. What you thought was secondary might be the actual hook. What you thought was self-evident might be invisible. What you thought was clever might reduce clarity. What you thought needed long explanation might only need six better words.

This is why speed matters, but not in the shallow startup cliché sense of “move fast” as a personality trait. Speed matters because delayed learning is expensive. Every extra week spent polishing an untested page is a week you are not learning from reality.

That is a much better signal because it tells you whether your assumptions were right, but usually they were not.

A more practical way to build pages now

The process I increasingly trust is much less romantic and much more useful.

Step 1: Start with existing truth, not blank-page genius

Do not begin by trying to “write the site.”

Start by assembling raw materials:

• sales calls
• voice notes
• service descriptions
• proposals
• onboarding docs
• objection handling
• client results
• old pages
• offer decks
• FAQs
• internal notes on who the offer is for and not for

That material usually contains the real language already. Not in polished form, but in substance.

This matters because blank-page writing encourages performance. Source material encourages specificity.

Step 2: Use AI to force draft velocity

Feed the material in and get a draft. Not because the draft is sacred, but because it exists and existence matters. A weak draft can be edited. A nonexistent page cannot be tested.

At this stage, I am mainly looking for:

• structure
• missing sections
• obvious repetition
• rough headline directions
• candidate framing angles
• useful phrases worth keeping

I am not asking, “Is this publish-ready?” I am asking, “Do I now have a thing I can react to?” That is enough.

Step 3: Edit for hierarchy before polish

This is where a lot of people waste time. They start polishing lines before they have fixed the message order.

The first question is not whether a sentence sounds elegant. The first question is whether the right idea appears in the right place at the right level of emphasis.

I usually want to know:

• What is the page trying to do?
• Who is it for?
• What must they understand in the first five seconds?
• What must they believe before they click?
• What objections are predictable?
• What proof is necessary?
• What can be cut without loss?
• Is the main point actually in the largest text?

That is hierarchy work, and it matters more than polish.

Step 4: Publish before you feel fully ready

This is the painful one.

There is almost always a moment where the page feels 80% there, and your instinct is to spend another week trying to make it 95%. Sometimes that is correct. Often it is just fear wearing a quality costume.

If the page is clear enough, credible enough, technically functioning, and connected to some form of analytics, it is usually time to publish.

Because now the page can start teaching you.

Step 5: Treat analytics as editorial input, not just reporting

This is where the practical value really kicks in.

Do not just “check analytics.” Use them to generate editing hypotheses.

For example:

• High bounce rate may mean the initial framing is weak or mismatched to the traffic source.
• Low scroll depth may mean the above-the-fold section is not earning continued attention.
• Good traffic but weak conversion may mean the offer, CTA, or proof structure is off.
• Heavy scrolling but no action may mean people are interested but unconvinced.
• Good engagement on one page section may mean that section should move higher.
• Repeated questions from prospects may mean the page is not actually communicating what you think it is.

Analytics do not tell you the answer automatically, but they help narrow the field of possible problems.

That is much better than editing blindly.

Step 6: Iterate with hypotheses, not vibes

This is another important shift. Do not just keep “improving the copy” in a vague way. Make specific bets.

For example:

• “I think the current headline is too abstract.”
• “I think people do not understand what the offer actually is.”
• “I think the CTA is too soft for high-intent visitors.”
• “I think the case study proof is too buried.”
• “I think we are leading with what I care about, not what they care about.”

Then change one meaningful thing at a time where possible.

That does not need to become enterprise-grade testing theatre. But even a modest hypothesis-driven approach is better than endless instinctive tinkering.

The hidden skill shift: from writer to editor to operator

The more I do this, the less I think the premium skill is “being able to write perfect copy from scratch.” That is still useful. But it is no longer the main bottleneck.

The more valuable stack now looks something like this:

1. Taste

• Can you tell the difference between generic language and precise language?
• Can you feel when something sounds impressive but says nothing?
• Can you identify when a page is over-explaining, under-explaining, or speaking in the wrong register?

2. Editing judgment

• Can you cut ruthlessly?
• Can you preserve substance while improving clarity?
• Can you detect what is structurally wrong, not just grammatically weak?
• Can you simplify without flattening?

3. Message architecture

• Can you arrange ideas in the right order?
• Can you align page structure with buyer psychology?
• Can you determine what must be said early, what can wait, and what should disappear entirely?

4. Operational follow-through

• Can you get the page actually shipped?
• Can you gather inputs, coordinate assets, connect analytics, publish updates, and maintain iteration cycles without everything dissolving into “I’ll come back to it later”?
• This is where a lot of the new leverage sits.
• Once drafting becomes cheaper, orchestration becomes more valuable.

The vulnerability in this for me personally

What has been uncomfortable for me is realizing how much of my older process was bound up with control.

If I spent long enough on the copy, I could pretend I was reducing risk. If I obsessed over wording, I could feel like I was doing serious work. If I delayed launch, I could preserve the possibility that the site would succeed perfectly once it finally went live.

That fantasy is seductive. Once something is published, reality gets a vote, and reality is not always flattering.

Sometimes the page you worked hardest on does nothing. Sometimes the section you barely thought about ends up doing more work than the one you laboured over. Sometimes the market reveals that you were solving the wrong messaging problem entirely.

That is uncomfortable, but it is also the only way to get better.

I think that is the real gift here. AI lowers the emotional and operational cost of being wrong. If it is cheaper to be wrong, it becomes easier to learn.

That is incredibly valuable.

This is not an argument for low standards

I am not arguing for slop. I am not arguing that strategy no longer matters. I am not arguing that all writing is interchangeable. And I am definitely not arguing that AI-generated text should just be copied and pasted without thought.

Bad taste at high speed is still bad. Generic language is still generic. Weak positioning is still weak positioning. Pages that feel sterile, inflated, or contextless still underperform.

What I am arguing is narrower and more practical

You should stop treating private perfection as a substitute for market contact. You should stop assuming that longer pre-launch deliberation produces better commercial truth. You should stop putting all the weight on first publication, as if the page cannot be improved once it is live.

In most cases, the better process is:

• get the real inputs together
• generate a strong draft quickly
• edit for structure and clarity
• publish
• measure
• revise based on evidence

That is not lowering the standard. That is moving the standard closer to reality.

What I would tell anyone building a site right now

If I had to reduce this to a practical recommendation, it would be this:

Do not make the mistake of thinking the main job is writing the perfect site. The main job is building a page system that can learn.

That means:

• the message needs to be clear enough to test
• the page needs to be live
• the CTA needs to function
• the analytics need to be there
• the offer needs to be legible
• the editing process needs to continue after launch

That is the actual work.

The romantic image of the solitary genius crafting immaculate homepage copy in private is becoming less economically relevant. The useful person now is the one who can extract truth from messy inputs, produce a coherent draft quickly, improve it with judgment, publish it, and keep refining it as evidence comes in.

That is a stronger skill set.

And frankly, it is a more honest one.

The real bottleneck was never writing

For me, the biggest realization is that the bottleneck was not really copy quality. It was the cost of going live.

The emotional weight of trying to say everything exactly right. The fear of publishing something that felt incomplete. The illusion that one more round of refinement would finally remove uncertainty.

It never did.

Because uncertainty is not removed by polishing in private. It is reduced by exposure to reality. That is why taking the bottleneck out of going live is so valuable. That is why seeing the analytics matters. That is why putting some distance between your identity and your first draft is priceless.

That is why I increasingly think the real upgrade is not becoming a better first-draft copywriter.

It is becoming a better editor, a better systems thinker, and a better operator.

Once you see that, the question changes. Not: “How do I make this perfect before anyone sees it?” But: “How quickly can I make this clear enough to learn from?”

That question has done more for me than all the agonising ever did.

The goal is not private perfection. The goal is a page that can learn.

Practical checklist: what I would actually do

Before drafting

• Gather sales notes, docs, decks, FAQs, proposals, and voice notes.
• Identify the core offer, audience, and CTA.
• Write down the top three objections and top three proof points.

During drafting

• Use AI to create structure and options, not final truth.
• Generate multiple headline directions.
• Draft the page in sections rather than trying to “write the site” in one pass.

During editing

• Fix hierarchy first.
• Make sure the biggest text says the most important thing.
• Remove internal jargon unless the audience genuinely uses it.
• Cut anything that only exists to make you feel accurately represented.
• Keep anything that improves buyer comprehension.

Before publishing

• Check mobile.
• Check CTA links.
• Check form flow.
• Add analytics.
• Make sure the page answers: what is this, who is it for, why trust it, what do I do next?

After publishing

• Review engagement and conversion patterns.
• Note where people drop off.
• Collect repeated questions from prospects.
• Revise one meaningful thing at a time.
• Keep a running log of message hypotheses and what changed.

Final thought

I still care deeply about language. I still want my site to sound like me. I still want it to be sharp, coherent, and true to the work.

But I no longer believe that the highest expression of care is spending three months trying to get it perfect in private.

Now I think the higher form of care is this:

Get it live. Let reality speak. Then get better on purpose.