I use Claude for the repeatable parts of SEO: drafting schema markup, structuring content briefs and generating title tag options. I don't use it for competitive data, prioritization or anything that depends on knowing a client's business. Every piece of AI output gets reviewed before it reaches a live site or a client.
Where AI fits in my SEO work
I work with US service businesses and B2B brands on SEO, and Claude is part of my working week. A lot of the talk about AI replacing SEO work doesn't match how search works, so this is a plain account of where it helps and where it doesn't.
Schema markup
Schema markup is structured data that tells Google explicitly what a page is about. Writing it by hand is tedious and easy to get wrong, so I have Claude draft the JSON-LD from a short description of the page.
The prompt I use: "Generate JSON-LD schema for a [LocalBusiness / ProfessionalService / Article]. The business name is [X], located at [address], serving [service area], offering [primary services]. Include all required properties and common recommended properties."
The draft comes back in seconds. I check it against the client's real details, then validate it in Google's Rich Results Test and the Schema Markup Validator before it goes live. A slow, fiddly job becomes a few minutes of checking.
Work With John
If your site isn't bringing in enquiries, that's fixable.
I work with US service businesses and B2B brands on SEO, websites and lead follow-up. On a free strategy call, I'll tell you what's holding your site back, without a sales pitch.
Book a free strategy callContent briefs
A content brief sets out what a piece needs to cover: the target keyword, the search intent, the questions to answer, the structure and what competitors already cover. Written from scratch, a good brief takes most of an hour.
Once I have the keyword data and competitor research, I have Claude draft the structure. The prompt: "Create a content brief for an article targeting [keyword]. The target audience is [description]. The search intent is [informational/decision/local]. Key competitors ranking for this term cover [X, Y, Z]. The brief should include suggested H2 headings and the key questions to answer."
I then adjust the draft with what the AI can't know about the client, and the writer gets a brief with most of the structure in place.
Title tag options
When a page ranks but few searchers click, I want several title options to choose from. Claude generates 10 to 15 variations quickly, and I pick the two or three worth testing.
The prompt: "The current title tag is [X]. The target keyword is [Y]. The page ranks at position [Z] with a click-through rate of [%]. Generate 10 alternative title tags that match the search intent and are short enough to show in full on a phone. No clickbait, and no generic words such as 'best' or 'top'."
The AI supplies variety. I choose based on what the searcher wants.
Where I don't use AI
Competitive analysis needs real ranking data, real link metrics and real traffic estimates. A language model doesn't hold current search data, and when asked for it, it can invent plausible numbers. I use Ahrefs and Search Console for this.
Prioritization is a judgment call. Which of 40 technical issues to fix first depends on the site, its competitors and the business goals, and a generic ranked list from an AI isn't the judgment a client is paying for.
Client-specific content needs context the AI doesn't have: the client's voice, their results and their positioning. Without that context, AI writing comes out generic, and generic content ranks poorly and converts worse.
The workflow that holds up
A person sets the strategy, the AI handles repeatable tasks within it, and a person reviews and edits everything before it touches a live site or reaches a client.
That saves me real time on structural work each week. The judgment, the competitive analysis and the client-specific work take as long as they always did.
Anyone who says AI can replace the strategic side of SEO has either not tried it or is selling an AI tool. AI saves time on the tactical work, and the strategy still needs someone who knows what they're doing.
One habit worth adding: AI drafts carry recognizable tells, such as overused em dashes, forced lists of three and generic conclusions, that undercut otherwise good content. I built an open-source Claude skill that catches and fixes them, and it fits into this same workflow.
