Should you use AI to write your website copy? The honest answer

A Bellows Falls contractor spent $800 in late 2023 on an AI copywriting service that promised to "generate SEO-optimized website content in minutes." The service produced a new homepage, three service pages, and five blog posts. The text was grammatically correct, flowed smoothly, and hit all the right keywords ("trusted contractor," "quality craftsmanship," "customer-focused solutions," "serving Bellows Falls and surrounding areas with excellence"). The contractor published it, waited three months, and saw zero change in traffic or leads. When he tested his site in ChatGPT with the query "Who are good general contractors in Bellows Falls Vermont?" ChatGPT did not mention him. It cited three competitors instead.

He read the competitors' sites. They were written in plain language, included specific facts (years in business, project counts, pricing ranges, timelines), and felt like a real person explaining their business. His AI-generated copy sounded polished but said nothing. No pricing, no timelines, no specific services, no examples. Just vague, professional-sounding marketing language that ChatGPT and Perplexity had nothing to extract and cite. The AI-written copy was invisible to AI search because it contained no facts, only fluff.

This is the trap most businesses fall into when they use AI to write website copy. They ask ChatGPT or Claude to "write homepage copy for a general contractor in Bellows Falls Vermont," and the AI produces smooth, professional-sounding text that is generic enough to fit any contractor in any town. The copy is not wrong, but it is not useful. It does not answer questions, does not provide specifics, and does not differentiate the business from competitors. Google can detect that the copy is generic (and likely AI-generated), and AI search engines ignore it because there are no extractable facts.

The honest answer to "should you use AI to write your website copy?" is: yes, but not the way most people are using it. AI is excellent at producing a first draft, organizing ideas, and overcoming blank-page paralysis. But AI is terrible at adding the specifics, the voice, and the facts that make copy credible and useful. If you publish AI-generated copy as-is, you will sound like every other business that did the same thing, and you will be invisible to both human readers and AI search engines. If you use AI to draft the structure and then rewrite it with your voice, your facts, and your examples, you will produce better copy faster than writing from scratch.

This post explains why pure AI-generated copy fails (for both human readers and AI search), why pure human-written copy is slow and expensive, and how the hybrid approach (AI for structure, you for substance) gives you the best of both worlds. It also walks through the step-by-step process for using AI as a drafting tool without letting it turn your website into generic marketing fluff.

Why AI-generated marketing copy is easy to spot (and getting easier)

AI language models (ChatGPT, Claude, Gemini, and the dozens of AI copywriting tools built on top of them) are trained on massive datasets of web text. They learn patterns, phrases, and structures that appear frequently in their training data. When you ask an AI to write marketing copy, it reproduces the patterns it has seen most often.

The problem: marketing copy on the web is notoriously generic and formulaic. Phrases like "we are committed to excellence," "trusted expertise," "customer-focused solutions," "serving [location] and surrounding areas," "quality service you can count on," and "your satisfaction is our priority" appear on thousands of contractor websites, law firm sites, accounting sites, and service business sites. AI absorbs these patterns and reproduces them because they are common.

The result: AI-generated marketing copy sounds professional and polished, but it is instantly recognizable as AI-generated (or at least as generic marketing copy) because it uses the same phrases, the same structure, and the same vague language that every other AI-generated site uses.

Here are the telltale signs of AI-generated copy:

Excessive use of generic superlatives. "Trusted," "experienced," "professional," "reliable," "quality," "best-in-class," "industry-leading," "customer-focused." These words appear in nearly every AI-generated marketing paragraph because they are statistically common in marketing copy. Humans writing copy use these words occasionally, but AI uses them constantly.

Vague location phrasing. "Serving [town name] and surrounding areas" or "proudly serving [region name]" instead of listing specific towns. AI defaults to vague phrasing because it does not know which towns you actually serve. A human writing copy will list the towns by name ("We serve Brattleboro, Putney, Bellows Falls, Wilmington, and Westminster").

No specific numbers or facts. AI-generated copy rarely includes pricing, timelines, project counts, or years in business unless you explicitly provide those facts in your prompt. A human writing copy naturally includes specifics because they have the context. AI defaults to vague phrasing ("competitive pricing," "timely service," "years of experience") because it does not have the data.

Smooth but empty sentences. "We understand that every project is unique, and we tailor our approach to meet your specific needs." This sentence is grammatically perfect and sounds professional, but it says nothing. Every business could say this. AI produces these sentences because they are common filler in marketing copy.

Overuse of "we" and "our." AI-generated copy tends to be heavily self-referential ("We provide... we believe... our team... our commitment..."). Human-written copy tends to address the reader more directly ("You need... your project... here's what you can expect...").

No personality or voice. AI defaults to a neutral, professional tone because it has no personality. It cannot be funny, sarcastic, blunt, or conversational unless you explicitly prompt it to be (and even then, the results are often forced). Human-written copy has a natural voice that reflects the person or business writing it.

A Putney lawyer used ChatGPT to write his practice area pages in early 2024. He prompted: "Write a page about estate planning services for a Vermont lawyer." ChatGPT produced 400 words of smooth, professional copy: "Estate planning is an important step in securing your family's future. Our experienced estate planning attorneys understand the complexities of Vermont estate law and are committed to providing personalized, client-focused solutions. We work closely with you to create a comprehensive estate plan tailored to your unique needs and goals. Whether you need a simple will or a complex trust, our team has the expertise to guide you through the process with professionalism and care. Contact us today to schedule a consultation and take the first step toward peace of mind."

The lawyer published it. Three months later, he tested his site in ChatGPT and Perplexity with queries about estate planning lawyers in Vermont. He was not cited. He read competitors' pages. One competitor's page said: "Estate planning includes drafting a will, setting up trusts, naming guardians for minor children, designating health care proxies, and planning for estate tax. A simple will costs $500 to $800. A will with trusts typically costs $1,500 to $3,000. The process takes 2 to 4 weeks from initial consultation to signed documents. We have completed estate plans for over 200 Vermont families since 2010." The competitor's page was full of facts. The AI-generated page was full of fluff. ChatGPT cited the competitor because there was something to extract and cite. The lawyer rewrote his page with specifics (services, pricing, timeline, experience) and was cited two weeks later.

Why Google and AI search engines are devaluing AI-generated content

Google's March 2024 Helpful Content Update explicitly targeted low-quality, AI-generated content. Google's guidance: "Our focus continues to be on the quality of content, rather than how it was produced. That said, we recognize that AI has the ability to be a useful tool for content creation, and it also has the potential to be misused to create low-value content at scale."

Google is not penalizing all AI-generated content. It is penalizing AI-generated content that is generic, low-value, and created at scale to manipulate rankings. The signals Google uses to detect this type of content include:

Lack of first-hand experience or expertise. Google's E-E-A-T framework (Experience, Expertise, Authoritativeness, Trustworthiness) prioritizes content that demonstrates real experience and expertise. AI-generated content rarely demonstrates first-hand experience because the AI has not done the work. A contractor writing about a challenging remodel can describe specific problems encountered and how they were solved. AI writing about the same topic can only produce generic advice.

Generic, templated language. Google's algorithm is trained to recognize patterns. If your content uses the same phrases and structure as thousands of other sites (because you all used the same AI tool with similar prompts), Google recognizes the pattern and deprioritizes the content as low-value.

No unique value or insight. Google's Helpful Content Update prioritizes content that provides unique value (new information, detailed explanations, specific examples, expert insight). AI-generated content often rehashes existing information without adding anything new.

AI search engines (ChatGPT, Perplexity, Claude) devalue AI-generated content for a different reason: there is nothing to extract and cite. AI search works by reading content, extracting facts, and synthesizing answers. If your content is vague and generic (because you published AI-generated copy as-is), the AI has no facts to extract. It moves on to competitors with factual, specific content.

This creates a vicious cycle: you use AI to generate generic copy, Google deprioritizes it because it is low-value, and AI search ignores it because it is not extractable. You spent money on the AI tool, but you are less visible than before.

A Wilmington accountant used an AI copywriting service to generate blog posts in 2023. The service produced 30 posts (500 words each) on generic accounting topics ("the importance of bookkeeping," "tax planning tips for small businesses," "why you need a CPA"). The posts were grammatically correct and read smoothly, but they contained no Vermont-specific information, no specific tax numbers or dates, no examples from the accountant's practice, and no unique insight. Google indexed the posts but did not rank them (they appeared on page 5 or later for most queries). ChatGPT and Perplexity never cited them because there were no extractable facts. A competitor published 8 blog posts (1,500 to 2,000 words each) written by the CPA herself, with Vermont tax law specifics, detailed examples, and timelines. Google ranked the competitor's posts in the top 5 for most queries. ChatGPT and Perplexity cited them frequently. The accountant deleted the 30 AI-generated posts and started writing her own (using AI as a drafting tool, then rewriting with specifics). Her visibility improved within 8 weeks.

The three things AI cannot do (and why you still need a human in the loop)

AI language models are impressive, but they have clear limitations when it comes to writing website copy for a specific business. Here are the three things AI cannot do without human input:

AI cannot know your specific facts

AI does not know how long you have been in business, how many projects you have completed, what you charge, how long your projects take, which towns you serve, what licenses or certifications you hold, or what makes you different from competitors. Unless you provide these facts in your prompt, AI will default to vague language ("experienced," "competitive pricing," "serving the area").

Even if you provide some facts in your prompt, AI cannot verify them or add context. If you tell the AI "I've been in business for 18 years," the AI will include that fact, but it will not add "I've completed 240+ projects" or "I was one of the first contractors in the area to specialize in net-zero construction" unless you explicitly provide those details.

A human writing copy knows the context and naturally includes specifics. AI needs to be told every single fact, and even then, it often buries the facts in generic sentences.

AI cannot replicate your voice or personality

Every business has a voice (formal or casual, technical or plain-language, serious or humorous, blunt or diplomatic). AI defaults to a neutral, professional tone because it has no personality. It cannot replicate your voice unless you train it extensively (which takes as much time as writing the copy yourself).

If you are a Brattleboro contractor who talks to customers in a straightforward, no-BS way ("Here's what the project costs, here's how long it takes, here's what can go wrong, and here's how we handle it"), AI will produce copy that sounds like a corporate brochure ("We are committed to transparency and excellence in every project we undertake"). The voice mismatch is jarring, and customers notice.

A human writing copy (or heavily editing AI-generated copy) can inject personality, humor, or bluntness that makes the copy feel real and trustworthy. AI cannot do this on its own.

AI cannot provide examples or tell stories

The most credible, engaging website copy includes examples and stories. A contractor describing a challenging renovation in Newfane (what went wrong, how they solved it, what the outcome was). A lawyer describing a complex estate case (anonymized) and how they navigated it. A CPA walking through a real tax planning scenario for a Vermont small business owner.

AI cannot provide these examples because it has not done the work. It can generate hypothetical examples if you prompt it ("Write an example of a kitchen remodel that went over budget"), but hypothetical examples feel fake because they are fake. Real examples feel real.

A human writing copy naturally includes real examples from their experience. AI needs to be told the example in detail, at which point you are essentially writing the example yourself and just having the AI format it.

The hybrid approach: AI for structure, you for substance

The most effective way to use AI for website copy is as a drafting and structuring tool, not as a final writer. Here is the hybrid process that works:

Step one: Use AI to generate an outline or first draft

Give the AI a detailed prompt that includes your service, your location, and the key points you want to cover. Ask it to generate an outline or a rough draft.

Example prompt for a Brattleboro HVAC contractor:

"Write an outline for a homepage for an HVAC contractor in Brattleboro, Vermont. The business has been in business for 22 years. Services include furnace repair, furnace installation, AC repair, AC installation, ductwork, and heat pump installation. Service area is Brattleboro, Putney, Bellows Falls, and Wilmington. Include sections for services, service area, pricing, timeline, credentials, and a call to action."

The AI will produce an outline (or a rough draft if you ask for one). The structure will likely be sound (headline, intro, service list, service area, pricing, credentials, CTA), but the content will be generic.

Step two: Rewrite each section with your specific facts

Go through the AI-generated draft section by section. Replace every generic phrase with a specific fact.

Generic AI version: "We offer a full range of HVAC services to keep your home comfortable year-round."

Your rewrite: "We provide furnace repair, furnace installation, AC repair, AC installation, ductwork, and heat pump installation. Most service calls are completed the same day. Furnace installations typically cost $4,500 to $7,000 and are completed in one day."

Generic AI version: "With years of experience serving the Brattleboro area, our team is committed to quality and customer satisfaction."

Your rewrite: "We have been in business in Brattleboro for 22 years and have completed 1,200+ service calls and 140+ system installations. Owner Mike Johnson is a licensed HVAC contractor in Vermont (license #3456) and a NATE-certified technician."

The structure stays the same (the AI gave you a good skeleton), but the substance is now yours (specific facts, your voice, your experience).

Step three: Add examples, stories, or process details

AI-generated drafts rarely include examples or process details. Add them yourself. A paragraph describing a typical furnace installation (what you do, how long each step takes, what the customer should expect). A sentence mentioning a recent project ("We recently installed a Mitsubishi heat pump system in a 2,200-square-foot home in Putney, reducing heating costs by an estimated 40% compared to the old oil furnace").

Examples and process details make your copy credible and useful. AI cannot provide them, so you have to.

Step four: Edit for voice and readability

Read the copy aloud. Does it sound like you, or does it sound like a corporate brochure? If it sounds generic or stiff, loosen it up. Replace "We are committed to providing exceptional service" with "We show up on time, we do the work right, and we clean up when we're done." Cut unnecessary words. Break long sentences into shorter ones. Make it conversational.

This step is critical. AI-generated copy tends to be wordy and formal. Editing for voice makes it sound human.

A Townshend landscaper used this process to rewrite his services page. He prompted ChatGPT: "Write a services page for a landscaping business in Townshend, Vermont. Services include mowing, trimming, mulching, spring and fall cleanups, and snow removal. Include pricing and service area." ChatGPT produced a 300-word draft with sections for services, service area, and contact. The draft was generic ("We offer professional landscaping services to keep your property looking its best year-round"). He rewrote it section by section, adding pricing (mowing starting at $40 per visit, mulching $80 per yard delivered and spread, spring cleanup $300 to $600 depending on property size), service area (Townshend, Newfane, Jamaica, Londonderry), timeline (mowing scheduled weekly or biweekly from May to October), and credentials (18 years in business, certified Vermont arborist, member of the Vermont Tree Care Association). He added a sentence about a recent project (a native plant garden he installed in Jamaica). The final version was 60% his rewrite, 40% AI structure. The copy was clear, factual, and credible. ChatGPT cited him when users asked about landscaping in the area. The AI gave him the structure, he provided the substance.

Why the hybrid approach works for both Google and AI search

The hybrid approach (AI for structure, you for substance) produces copy that works for traditional Google search and for AI search because it combines the best of both:

The AI provides structure and saves time. Generating a first draft or outline with AI takes 2 to 5 minutes. Writing the same outline from scratch takes 20 to 30 minutes. The AI handles the boring structural work (headings, flow, transitions) so you can focus on the substance.

You provide the facts that make the copy extractable. Google's algorithm and AI search engines prioritize factual, specific content. By rewriting the AI draft with your facts (pricing, timelines, towns, credentials, project counts), you make the copy extractable and citeable. AI-generated fluff gets ignored. Fact-rich copy gets cited.

You provide the voice that makes the copy credible. Humans reading your website can tell the difference between generic marketing copy and copy written by a real person with real experience. Voice and personality build trust. The hybrid approach lets you inject your voice while still benefiting from the AI's structural help.

The result ranks well in Google and gets cited by AI. Google rewards content that demonstrates expertise and provides unique value (you rewrote the AI draft with your specific facts and experience). AI search rewards content that is extractable and backed by third-party validation (you added facts AI can extract, and your third-party presence validates those facts). The hybrid copy works for both.

A Westminster real estate agent used the hybrid approach to rewrite her buyer and seller guide pages. She prompted Claude: "Write a guide page for home buyers in Westminster, Vermont. Include sections for the buying process, typical timeline, costs, and tips for first-time buyers." Claude produced a 600-word draft with good structure but generic content ("The home buying process can be complex, but with the right guidance, you can navigate it successfully"). She rewrote each section with Vermont-specific facts (median home prices in Westminster, typical closing costs as a percentage, average time from offer to close in the current market, Vermont first-time buyer programs). She added a paragraph about a recent buyer she helped (first-time buyers who purchased a farmhouse in Westminster, what challenges they faced, how she helped them navigate inspection issues). The final page was detailed, specific, and useful. Google ranked it in the top 5 for "buying a home in Westminster VT." Perplexity cited it when users asked about the home buying process in southern Vermont. The AI gave her the structure in 3 minutes, she spent 45 minutes adding substance, and the result outperformed anything she could have written from scratch in the same time.

How to prompt AI to get a useful first draft (not generic fluff)

If you are going to use AI as a drafting tool, the quality of your prompt determines the quality of the output. A vague prompt ("Write homepage copy for a contractor") produces generic fluff. A detailed prompt with specific instructions produces a useful first draft you can build on.

Here is a framework for writing effective prompts:

Include your service, location, and key differentiators

Tell the AI exactly what you do, where you do it, and what makes you different. The more specific your prompt, the less generic the output.

Vague prompt: "Write homepage copy for a lawyer."

Better prompt: "Write homepage copy for an estate planning and family law attorney in Brattleboro, Vermont. The attorney has been practicing for 18 years, is a member of the Vermont Bar Association, and offers flat-fee pricing for most services. Service area is Brattleboro, Putney, Bellows Falls, and Wilmington."

The better prompt gives the AI context, so the output will be more specific (though you will still need to rewrite it with your exact pricing and process details).

Request structure, not fluff

Ask the AI to organize information, not to write persuasive marketing copy. Prompts that ask for "compelling," "engaging," or "persuasive" copy tend to produce generic superlatives and empty phrases. Prompts that ask for "clear," "factual," or "structured" copy tend to produce more useful output.

Fluff-inducing prompt: "Write compelling homepage copy that will engage visitors and convince them to contact us."

Better prompt: "Write a clear, factual homepage that includes sections for services offered, service area, typical project costs and timelines, credentials and experience, and a call to action."

The better prompt focuses on information architecture, not persuasion. You can add persuasive elements later if needed, but the foundation will be factual and structured.

Provide the facts upfront

If you have specific facts (years in business, pricing, certifications, project counts), include them in the prompt. The AI will incorporate them (though you should still double-check the output).

Prompt without facts: "Write a services page for an HVAC contractor."

Prompt with facts: "Write a services page for an HVAC contractor who has been in business for 22 years, is a licensed contractor in Vermont (license #3456), and offers furnace installation ($4,500 to $7,000), AC installation ($3,800 to $6,200), and service calls (starting at $95). Service area is Brattleboro, Putney, and Bellows Falls."

The second prompt gives the AI the facts, so the output will include them. You will still need to edit for voice and add examples, but the factual foundation will be there.

Ask the AI to avoid marketing clichés

You can explicitly instruct the AI to avoid generic phrases. This does not always work perfectly (AI models have strong learned patterns), but it helps.

Example instruction: "Write a homepage for a general contractor in Bellows Falls, Vermont. Avoid generic phrases like 'trusted expertise,' 'committed to excellence,' 'customer-focused,' and 'quality service.' Focus on specific facts like services offered, years in business, pricing ranges, and project timelines."

This prompt steers the AI away from fluff and toward substance. The output will still need editing, but it will be less generic than a prompt without this instruction.

A Grafton financial planner tested different prompts to draft a retirement planning page. First attempt (vague prompt): "Write a page about retirement planning services." The output was 400 words of generic advice ("Retirement planning is essential to securing your financial future. Our experienced advisors work closely with you to create a personalized plan..."). Second attempt (detailed prompt): "Write a retirement planning services page for a CFP in Grafton, Vermont. Include sections for: what retirement planning includes (401k rollovers, IRA contributions, Social Security timing, withdrawal strategies), typical costs (fee-only planning at $150/hour or $2,000 flat fee for a comprehensive plan), timeline (initial plan takes 4 to 6 weeks), and credentials (CFP designation, 16 years of experience, member of NAPFA). Avoid marketing clichés and focus on factual descriptions." The second output was much better. It had clear sections, included the facts from the prompt, and avoided most of the fluff. The planner spent 30 minutes editing it (added examples, tightened the language, adjusted the voice), and the result was a strong, factual page that ChatGPT cited when users asked about retirement planning in Vermont.

The five places you should never use AI-generated copy as-is

Even with the hybrid approach, there are certain types of content where AI-generated copy is particularly weak and where you should do most of the writing yourself:

Your About page or bio

Your About page needs personality, story, and credibility. AI cannot tell your story (it does not know your story), and AI-generated bios sound generic and corporate. You can use AI to draft a structure ("Write an About page outline for a contractor: sections for background, experience, credentials, and values"), but you need to rewrite every section with your real story.

Example: AI might write "I have always been passionate about construction and committed to delivering quality results for every client." You should rewrite it as "I started framing houses in the summer after high school and never left the trades. Twenty-two years later, I've built or remodeled over 200 homes in Windham County, and I still get excited about a clean miter joint and a square corner."

Case studies or project examples

AI cannot write case studies because it does not know your projects. You can prompt AI to create a template ("Write a case study template for a home remodel project: sections for client need, challenges, solution, and outcome"), but you have to fill in the actual project details yourself. Do not let AI invent hypothetical examples (they will sound fake).

Detailed process explanations

AI can give you a generic process outline ("Consultation, proposal, contract, work begins, completion"), but it cannot describe your actual process with the insider details that make it credible. A Brattleboro contractor explaining their remodeling process might write: "After we finalize the design and sign the contract, we pull permits (this usually takes 1 to 2 weeks in Brattleboro, longer in some surrounding towns). Then we order materials (cabinets and countertops have the longest lead times, typically 4 to 6 weeks). We schedule the project to start once materials arrive and permits are approved." Those details come from experience. AI cannot provide them.

FAQ answers that require local or specialized knowledge

AI can generate generic FAQ questions ("What does estate planning cost?" "How long does a kitchen remodel take?"), but the answers need to be specific to your business, your location, and your expertise. If you are a Vermont lawyer writing an FAQ answer about estate taxes, you need to include Vermont-specific estate tax thresholds and how they differ from federal thresholds. AI will give you federal information, but it will not know the Vermont specifics unless you feed them into the prompt (at which point you are writing the answer yourself).

Anything where credibility and trust are critical

If you are writing content for a YMYL (Your Money or Your Life) topic (health, legal, financial), AI-generated copy is particularly risky because credibility and E-E-A-T are mandatory. A CPA writing about tax strategies, a lawyer writing about estate planning, or a doctor writing about treatment options cannot rely on AI to get the facts right or to demonstrate expertise. These pages need to be written (or very heavily edited) by the expert, with citations and credentials clearly displayed.

A Putney therapist tried using AI to write her services page (therapy approaches, what to expect, pricing). The AI-generated copy was smooth but generic ("I offer evidence-based therapy in a supportive, client-centered environment"). She rewrote it herself: "I am a Licensed Clinical Social Worker (LICSW) in Vermont with 14 years of experience treating anxiety, depression, and trauma. I am trained in Cognitive Behavioral Therapy (CBT) and EMDR. Sessions are 50 minutes and cost $140 (I accept BCBS and MVP insurance, and I offer a sliding scale for clients without insurance). I typically see clients weekly for the first 8 to 12 weeks, then biweekly or monthly as symptoms improve. Most clients notice meaningful improvement within 3 to 4 months." The rewrite included credentials, specifics, and process details that built trust. The AI draft had none of that. For YMYL content, the expert needs to write or heavily edit the copy.

How to test whether your copy is AI-detectable (and what to do if it is)

If you published AI-generated copy and you want to know whether it is detectable as AI-written, here are three quick tests:

Test one: Read it aloud

AI-generated copy often sounds smooth but generic when you read it aloud. If you find yourself thinking "this could be any business" or "this sounds like a brochure," it is probably AI-detectable. Human-written copy tends to have rhythm, personality, and specifics that make it feel real.

Test two: Highlight the facts

Go through your copy and highlight every specific fact (a number, a name, a credential, a timeline, a price, an example). If you have fewer than one fact per paragraph, your copy is too vague (whether it is AI-generated or human-written). AI-generated copy tends to be fact-light because the AI does not know your specifics.

Test three: Run it through an AI detection tool

Tools like GPTZero, Originality.ai, and Copyleaks offer AI detection. They are not 100% accurate (they sometimes flag human-written text as AI, and they sometimes miss AI-generated text), but they give you a rough signal. If your copy scores as "likely AI-generated," it probably reads as generic and formulaic. Rewrite the flagged sections with specifics and personality.

What to do if your copy is AI-detectable:

Rewrite the vague sections with facts. Replace "we offer quality service" with "we complete most service calls the same day, and we offer a 2-year warranty on all installation work." Replace "serving the area for years" with "we have been in business in Brattleboro since 2007." Add examples, process details, and personality. The goal is not to trick the AI detector (that is a losing game), the goal is to make your copy factual, specific, and useful so that both human readers and AI search engines trust and cite it.

A Bellows Falls accountant ran GPTZero on her services page (which she had generated with an AI tool in 2023 and published as-is). GPTZero flagged it as 92% likely AI-generated. She read it and realized it was full of generic phrases ("comprehensive accounting services," "personalized attention," "trusted expertise"). She rewrote it with specifics: services offered (tax prep, bookkeeping, business consulting, IRS audit support), pricing (tax prep starting at $200, bookkeeping $150 to $400 per month depending on volume), credentials (CPA license #5678, member of Vermont Society of CPAs, 19 years in practice), and service area (Bellows Falls, Rockingham, Springfield, Chester, Westminster). She ran GPTZero again. Now it scored as 15% likely AI-generated (in the range of human-written content with some formulaic phrasing, which is fine). More importantly, ChatGPT started citing her when users asked about CPAs in the area because her rewritten page was full of extractable facts.

Why "AI-assisted" disclosure does not help (and might hurt)

Some businesses add a disclaimer to their AI-generated content: "This page was written with the assistance of AI" or "Content on this site may include AI-generated text." The intent is transparency, but the effect is counterproductive.

It signals to readers that the content might be generic or unreliable. If you tell a potential customer "this was written by AI," they immediately question whether the content reflects real expertise or experience. Even if the content is good, the disclosure undermines trust.

It does not help with Google or AI search visibility. Google does not give you credit for disclosing AI use. If the content is low-quality and generic, the disclosure does not save you from being deprioritized. If the content is high-quality and factual, you do not need the disclosure.

It opens you up to scrutiny. If a reader or competitor sees the disclosure, they may look more critically at your content to see if it is generic or inaccurate. You are inviting skepticism.

The better approach: use AI as a tool (just like you use spell-check, grammar tools, or content management systems), but take full ownership of the final content. If you rewrote the AI draft with your facts, your voice, and your examples, the content is yours. You do not need to disclose that you used AI to generate a first draft any more than you need to disclose that you used Microsoft Word to type it.

A Wilmington retail shop owner added an AI disclosure to her website footer in early 2024 ("Some content on this site was created with the assistance of AI language models"). A customer emailed asking whether the product descriptions were accurate or just AI-generated placeholders. The owner realized the disclosure was creating doubt rather than building trust. She removed the disclosure, rewrote the product descriptions with her own voice and details, and the issue disappeared. The lesson: if your content is good, you do not need to disclose AI use. If your content is not good, the disclosure does not fix it.

The four rules for using AI to draft website copy without sounding like everyone else

If you want to use AI as a drafting tool (which is smart and efficient), follow these four rules to avoid the generic AI voice:

Rule one: Always rewrite with your specific facts

Never publish an AI draft as-is. Go through every section and replace vague phrases with specifics (pricing, timelines, towns, project counts, credentials, examples). The facts are what make your copy unique and useful.

Rule two: Edit for voice and personality

Read the draft aloud and ask: does this sound like me, or does it sound like a corporate brochure? If it is too formal or generic, loosen it up. Add contractions, break long sentences, use plain language, inject personality.

Rule three: Add examples and stories

AI cannot provide real examples or tell real stories. Add them yourself. A sentence about a recent project, a paragraph describing a typical client scenario, a case study with real outcomes. Examples make your copy credible and engaging.

Rule four: Focus on usefulness, not persuasion

The best website copy answers questions and provides information (what you do, how much it costs, how long it takes, what the process is). AI is tempted to write "persuasive" copy full of superlatives and vague claims. Push back against that. Make the copy useful first, persuasive second. Useful copy converts better because it builds trust.

A Londonderry electrician used these four rules to rewrite his services page. He prompted ChatGPT for a first draft, then rewrote it: replaced "competitive pricing" with "service calls start at $95, panel upgrades typically cost $1,800 to $3,500," replaced "serving the area for years" with "licensed electrician in Vermont for 24 years, completed 1,600+ service calls," replaced "we are committed to excellence" with "we show up on time, we explain what we find, and we do not sell you anything you do not need." He added an example (a recent panel upgrade in Weston that allowed the homeowner to add EV charging). The rewrite took 40 minutes. The result sounded like a real person explaining their business, not a corporate brochure. ChatGPT cited him when users asked about electricians in the area. The AI gave him the structure, the four rules gave him the substance.

What about using AI for blog posts, guides, or long-form content?

The hybrid approach works for blog posts and long-form content, but the balance shifts. For a homepage or services page (300 to 600 words), you can use AI to generate a draft and then rewrite 60% to 80% of it. For a 1,500 to 2,000-word blog post or guide, rewriting 60% to 80% is a huge amount of work (you are essentially writing the post yourself). The time savings shrink.

Here is how to use AI for long-form content without letting it take over:

Use AI to generate an outline. Prompt: "Write a detailed outline for a blog post about [topic] for [audience]. Include an introduction, 5 to 7 main sections with subheadings, and a conclusion." The AI will give you a solid structure. You write the content for each section yourself.

Use AI to draft individual sections, then rewrite them. If you are stuck on a section, prompt the AI to draft it, then rewrite the draft with your facts, examples, and voice. This is faster than staring at a blank page.

Use AI to generate FAQ questions or examples. If you are writing a guide and you need FAQ questions, prompt the AI: "Generate 10 frequently asked questions about [topic] for [audience]." Pick the best 5, then write your own answers. Or prompt the AI to generate a hypothetical example, then replace the hypothetical details with real details from your experience.

Never publish AI-generated long-form content as-is. The longer the content, the more obvious it is when AI wrote it. Generic 2,000-word blog posts are easy to spot (they are fact-light, full of fluff, and lack personality). If you are publishing long-form content, you need to write most of it yourself (or hire a writer who will).

A Brattleboro CPA wanted to publish a guide about Vermont small business taxes. She prompted Claude to generate an outline. Claude produced a 12-section outline covering entity types, tax obligations, deductions, estimated taxes, payroll, and more. She wrote each section herself (1,500 words total), adding Vermont-specific tax rates, deadlines, and examples from her client work. She used AI to draft one section (estimated quarterly taxes) that she was stuck on, then rewrote the draft with Vermont-specific details and a step-by-step process. The final guide was 90% her writing, 10% AI structure. The guide ranked well in Google and was cited by ChatGPT and Perplexity because it was detailed, Vermont-specific, and full of real examples. If she had published the AI draft as-is, it would have been generic federal tax advice that every CPA site already has, and it would have been invisible.

How to audit your existing website copy for AI-detectable fluff (and fix it)

If you published AI-generated copy in 2023 or 2024 and you are not sure whether it is helping or hurting you, run this audit:

Step one: Check your homepage, About page, and top three service pages

Read each page carefully. Highlight every sentence that contains a specific fact (a number, a credential, a price, a timeline, a town name, an example). Count the highlighted sentences. If fewer than 30% of your sentences contain facts, your copy is too vague (whether it is AI-generated or human-written).

Step two: Look for generic phrases

Search your pages for these phrases (use your browser's find function): "trusted," "experienced," "professional," "quality," "committed to excellence," "customer-focused," "serving the area," "years of experience" (without a number), "competitive pricing" (without actual pricing). Every instance of these phrases is a red flag. Rewrite them with specifics.

Step three: Test a sample in an AI detection tool

Copy your homepage text into GPTZero or Originality.ai. If it scores above 70% likely AI-generated, your copy is generic and formulaic. Rewrite the flagged sections with facts and personality.

Step four: Test your AI visibility

Run three queries about your business in ChatGPT and Perplexity. Are you cited? If no, read the competitors who are cited and compare their content to yours. What facts do they include that you do not? Pricing? Services? Towns? Credentials? Add those facts to your site.

Step five: Rewrite the worst pages first

Pick the two pages with the most generic copy (usually the homepage and a key service page). Rewrite them using the hybrid approach (keep the structure, replace every vague phrase with a fact, add examples, edit for voice). Test your AI visibility again two weeks later. If you are now being cited, continue rewriting other pages.

A Putney contractor ran this audit in mid-2024. His homepage had 18 sentences, only 4 of which contained facts (22%). His services page was worse (12 sentences, 2 facts). He searched for generic phrases and found 14 instances ("trusted contractor," "quality workmanship," "years of experience," "serving the area," "committed to excellence"). He ran his homepage through GPTZero (87% likely AI-generated). He tested ChatGPT (not cited). He rewrote his homepage and services page over two evenings (added service list, town list, pricing, years in business, project count, license number, two project examples). He tested ChatGPT two weeks later. Now he was cited in three out of four queries. The rewrite took 4 hours total and transformed his visibility. The generic AI copy was killing him, the factual rewrite saved him.

The bottom line: AI is a tool, not a writer

AI is excellent at generating structure, overcoming blank-page paralysis, and saving time on first drafts. AI is terrible at providing the facts, voice, and examples that make website copy credible, useful, and citeable. If you publish AI-generated copy as-is, you will sound like every other business that did the same thing, and you will be invisible to both Google and AI search engines. If you use AI as a drafting tool and then rewrite the output with your facts, your voice, and your examples, you will produce better copy faster than writing from scratch.

The businesses that win in 2026 are the ones that understand this balance. They use AI to save time on structure and organization. They do the hard work of adding substance (facts, examples, personality) themselves. The result is copy that is efficient to produce, credible to read, and extractable for AI search. The businesses that try to shortcut the process (publish AI copy as-is) will be invisible. The businesses that refuse to use AI at all will be slower and less efficient than competitors who use it strategically.

Use AI as a tool. Own the final product. Make it factual, specific, and useful. That is the honest answer.

Want to see whether your website copy is AI-detectable, get a page-by-page audit of vague phrases and missing facts, and get a rewrite checklist that makes your content extractable for AI search without sounding generic? The free 27-point local SEO audit now includes a copy quality diagnostic. We check your homepage and service pages for fact density, flag generic AI phrases, test your content in ChatGPT and Perplexity to see what gets extracted (or ignored), and give you a prioritized list of pages to rewrite with specific rewrite suggestions. No pitch, just the diagnostic and the roadmap.