AI models can research, outline and draft far faster than a person. They also state wrong things confidently, repeat the consensus of what already ranks, and write in a recognisable voice that readers skim past. Google's guidance is clear that how content is made matters less than whether it is helpful, and that mass-producing pages mainly to rank breaks its spam policies.
The practical answer is a process where AI does the parts it is good at and a person is responsible for the rest.
Short on time? Comppound does this work for SaaS companies. AI-native, and a person reviews every page.
Have us do itWhat AI does well
Use models for the work that is slow for people and easy to check:
- Summarising the top results and listing what they cover
- Turning a brief and your notes into a first draft
- Suggesting headings, questions and internal links
- Rewriting for clarity, length or reading level
- Producing variations of titles and descriptions to choose from
What a person must own
Some jobs cannot be delegated, because they are where pages go wrong:
- The angle: what this page says that others do not
- Facts, figures and claims about your product and competitors
- Proof only your company has: screenshots, customer stories, data
- Judgement calls: who a product suits, where the competitor wins
- The final read, as the reader would read it
Comppound can do this for you
Want the content produced for you?
Comppound writes the briefs, drafts with AI, adds the proof only your team has and has a person review every page before it goes live. You approve; we publish and keep it current.
- Briefs built from real searches
- Human review on every page
- A steady publishing rhythm, every month
Checks between draft and publish
Run every draft through the same checks before a person reads it, so their time goes on judgement rather than catching basics: unsourced figures, placeholder text, an opening that does not answer the search, missing links, and stock phrases that make text read as machine-written. Then the reviewer reads the whole page and signs it off by name.
Keep a voice
Write a short style guide: how you address the reader, words you use and avoid, how you handle competitors, what a good opening looks like. Give it to the model with every brief, and use it as the reviewer's checklist. Over time it is what makes your pages sound like one company.
Worked example · an illustration, not a client result
Tallyroom's drafting process
Tallyroom's founder reviews every page but writes almost none from scratch. The time saved went into gathering proof and reviewing more carefully.
| Step | Who | Output |
|---|---|---|
| Brief | Person | One-page brief with angle and proof |
| Research summary | AI | What the top results cover |
| First draft | AI, from the brief | Full draft with headings and links |
| Automatic checks | Tool | List of flagged issues |
| Review and edit | Person | Facts checked, proof added, signed off |
Do it yourself
Draft one page this way
- 1Write the brief from the previous lesson.
- 2Give the brief, your style guide and your notes to an AI model and ask for a draft.
- 3Run the draft through the draft quality check.
- 4Check every fact and claim against a source, and add your own proof.
- 5Read it start to finish as a buyer would, then publish with a named reviewer.
Free tool for this: Draft quality check, Readability checker
Common mistakes
Publishing unread drafts
Unreviewed AI drafts carry wrong facts and generic advice. One wrong claim about a competitor can cost far more than the time saved.
Scaling before quality
Producing many pages quickly from thin briefs is the pattern Google's scaled content abuse policy targets.
No style guide
Without one, every draft sounds like the model's default voice, and so does every competitor's.
Rather not do it yourself?
Want the content produced for you?
Comppound writes the briefs, drafts with AI, adds the proof only your team has and has a person review every page before it goes live. You approve; we publish and keep it current.
- Briefs built from real searches
- Human review on every page
- A steady publishing rhythm, every month
Questions
Google says it rewards helpful content however it is produced, and acts against content made mainly to manipulate rankings, including at scale. Quality and purpose are what count.
Many companies note how content is produced and who reviewed it. Naming the reviewer is the more useful signal for readers.
Key terms
- AI-Assisted WritingAI-assisted writing uses language models to help research, outline, draft or edit content. Google judges the result on quality, not on how it was made.
- AI-Generated Content PolicyAI-generated content policy covers the rules search engines and platforms apply to AI-made content, focused on quality, intent, spam and disclosure.
- Scaled Content AbuseScaled content abuse is producing many pages mainly to manipulate rankings rather than help users, whether by AI, people or templates.
- Editorial ReviewEditorial review is the checking of content before publication for accuracy, clarity, structure, style and fit with the brief, usually in several passes.
- Fact-CheckingFact-checking verifies every factual claim in content, including statistics, quotes, dates and product details, against primary sources before publishing.
- Style GuideAn editorial style guide sets the rules for how a company writes: spelling, grammar, terminology, formatting, tone and how to refer to the product.
- Brand VoiceBrand voice is the consistent personality in a company's writing, how it sounds across content, product and support, while tone adapts to each situation.
- HallucinationA hallucination is when an AI model states something false or unsupported as if it were fact, such as a wrong price, a fake feature or an invented source.
