Website copy: writing it yourself versus hiring

Copy is the part of a web project that decides whether it works. Design gets a visitor to stay for four seconds. Words are what make them call.

It is also the part that stalls. Almost every late website project is late because pages are sitting empty waiting on the owner, and the owner is avoiding them because writing about your own business is uncomfortable.

What makes copy good for a local business?

Specificity. That is nearly the entire answer.

"We provide quality service with attention to detail" describes nothing and could be pasted onto any business in any trade. "We can usually get a same-day appointment for a broken water heater, and we will tell you on the phone whether it is worth repairing" describes a real company that a real person can decide about.

The second thing is answering the questions people actually have. Do you serve my town. Do you handle my situation. What will this cost, roughly. How fast can you come. Who will show up at my house. Most sites answer none of those and instead explain that the company was founded in 1998.

The third is proof. Claims without evidence are noise, and the specific evidence that changes behavior is narrower than most owners assume. Which trust signals actually change behavior is worth knowing before you spend a paragraph on badges nobody reads.

If a sentence could appear on a competitor's site without changing a word, delete it.

When should I write it myself?

When you are the person who explains this work to customers all day, which is most owners.

The method that works is a recording, not a blank page. Have somebody ask you the five questions customers ask most, answer them out loud the way you would on the phone, and transcribe it. That transcript, lightly cleaned up, is better copy than most agencies will write, because it contains the details only you know.

Writing it yourself also has a second benefit that has nothing to do with money. You will discover which parts of your offer you cannot explain clearly, and those are usually the parts customers are confused about too.

The failure mode is polish. Owners write a good honest draft and then edit the life out of it, replacing plain sentences with formal ones, until it sounds like everyone else. Keep the plain version.

How many pages can one recording actually produce?

When is hiring the right call?

Three situations.

Volume. If your page map is fifteen pages, you are not going to write fifteen pages. You will write four and the project will sit. Paying somebody to do the other eleven is buying a launch date.

Consistency. If several people have written parts of your site over the years, it reads as several companies. A writer going through it once fixes that in a way piecemeal editing never does.

Search structure. Copy that has to rank for different terms in different places is a technical exercise, not only a writing one. Whitespark's Local Search Ranking Factors report, published 6 November 2025 with 47 local search experts scoring 187 factors, ranks keywords in the landing page title tag at 17th with 146 points, geographic keyword relevance of content at 18th with 142, and keywords in landing page headings at 20th with 135. That is expert opinion rather than a controlled test, and Whitespark says so, but it is a clear statement that near-identical city pages are not doing the job. The words on each page have to differ meaningfully, and why you rank in one part of town but not another is a problem that duplicated city copy makes worse rather than better.

What you should not hire out is the source material. A writer needs your actual details, your actual process, and your actual objections. If you hand a writer nothing, you get generic copy back and it is not their fault. What to prepare before a web project starts is largely a list of the raw material a writer or designer cannot proceed without.

Should I write any of this for AI search?

No, and this is worth knowing because it is currently the most oversold service in the category.

Google's own guidance on generative AI features names the tactics it says do not work, in its own mythbusting section: artificial content chunking, rewriting content specifically for AI systems, special machine readable files, and seeking inauthentic mentions. Its summary line is the useful one to remember: from Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO.

The academic work points the same way, and it is worth knowing what it actually tested. Aggarwal and colleagues published "GEO: Generative Engine Optimization" at KDD 2024, building a benchmark of 10,000 queries and simulating an AI answer engine: fetch the top five Google results, then feed that cleaned text to GPT-3.5 to synthesise a cited answer. Against a no-optimization baseline scoring 19.3 on their word-count metric, adding direct quotations scored 27.2, adding statistics 25.2, and citing sources 24.6. Keyword stuffing scored 17.7, below the baseline. It made things worse. And writing in a more authoritative tone produced no significant improvement at all, which the authors read as evidence that these systems already shrug off that kind of change.

Two cautions travel with that. It is a simulation over five fixed sources, not a live product, and nothing in the benchmark is a local commercial query, so it has never been validated for a plumber in San Jose. And the widely quoted "40% more AI visibility" figure is that 27.2 against 19.3, on one tactic, on a word-count metric, in a lab. It is not traffic.

So the practical translation is boring and it is the same advice as the rest of this article. Say specific things. Quote real people accurately. Cite where a number came from. Do not stuff keywords, and do not pay anybody to rewrite your homepage for a machine.

What about generating the copy?

The failure mode is the same one as hiring a writer with no brief, only faster and cheaper.

Text produced from a prompt reads as generic because it is assembled from generic material, and the specificity that makes local copy work is exactly the thing it does not have. It will confidently write that you are committed to quality. It cannot write that you will tell somebody on the phone whether a water heater is worth repairing, because it does not know that.

There is a length trap attached. Generated copy is long, and length has a cost. Google's mobile benchmarks research with SOASTA, trained on bounce and conversion data at 90% prediction accuracy, reports that as the number of elements on a page goes from 400 to 6,000, the probability of conversion drops 95%. Element count is not word count, but it is a decent proxy for clutter, and padded pages accumulate elements fast.

How do I work with a writer without wasting the budget?

Give them a transcript, not a brief. Thirty minutes of you talking is worth more than three pages of adjectives about your brand personality.

Give them the objections. The five reasons people do not hire you, in your own words. Good copy answers objections and a writer cannot guess yours.

Then review for accuracy first and style second. The most useful feedback is "we do not actually do that" and "customers never say it that way." The least useful is "make it pop." How you deliver that feedback determines how many rounds this takes, and using your revision rounds well is the difference between two rounds and seven.

And hand off the parts of the project that contain no judgment at all. Nobody needs the owner's opinion on a server configuration, and hosting decisions for a small business site should be made once by whoever is accountable for uptime. Save your attention for the words, because those are the part only you can supply.

Where does this break down?

Two places.

When you are not actually the explainer. Some owners are excellent operators and genuinely poor at describing the work out loud. If your recording produces eleven specifics instead of sixty-one, that is your answer and there is no shame in it. Hire, and spend your time on the objections list instead.

When the copy is not the constraint. If nobody is arriving, better words are read by nobody. Copy multiplies traffic. It does not create it, and doing them in the wrong order is how owners conclude that marketing does not work.

What to do this week

Record yourself answering five questions: what you do, who you do it for, what happens when someone calls, what it typically costs, and why customers choose you over the obvious alternative. Use your phone. Do not prepare.

Transcribe it. Most phones will do this automatically now.

Read the transcript and run the six step count above. The number of specifics you found is the honest measure of how much of the site you can write.

Then take one page, your most important service, and write it from those details. If that page comes out well and it took ninety minutes, you can write your own site. If you have not started it by next Friday, hire, and stop pretending otherwise.

Questions about your website copy?

Email me at eric@seod.com and paste in the text from your homepage, exactly as it stands. I will mark the sentences a customer would actually care about and the ones that could belong to any business in your category. It is usually a short and slightly uncomfortable reply, and it tells you immediately whether you need a writer.

There is more on running a project well in the web design and branding library.