SEO content strategy is mostly research and decisions, and both eat tabs. Before a single word gets written, someone has to read the live SERP for a keyword, judge the intent, size up who already ranks, decide whether the site can realistically compete, and turn all of that into a brief a writer can follow. Then it repeats for the next keyword, and the one after that, until the calendar is full and the actual strategy, the clusters and the internal links that compound over time, never gets the attention it deserves.
An AI agent fits the research layer cleanly. It searches the live SERP instead of guessing, so a brief is built on what Google is rewarding this week, not what ranked last year. It reads an uploaded sitemap and proposes internal links with real anchor text. It drafts titles and metas at the volume a content calendar actually demands. And because it runs on whichever model you pick, you can hand the same intro brief to Claude and to GPT and see which draft opens stronger before a writer ever touches it. That last part is the Keimodel angle: you choose the model on evidence, not habit.
This is the strategy and research layer, not generic blog writing. Set your site's facts, your domain strength, and your house style in Memory once, and every brief, cluster, and title below comes back grounded in your site instead of a blank one. You keep the judgment. The agent does the SERP-reading and the production.
Capabilities this leans on
Do this first. Every brief, cluster, and title below comes back grounded in your site and your house style instead of a generic one.
Remember these facts about my SEO and content work: I run content strategy for Bloomfield, a houseplant care brand at bloomfield.com that sells plants and care supplies and runs a large blog. Our audience is beginner to intermediate houseplant owners. Primary keyword themes are plant-care troubleshooting, watering and light guides, and plant buying guides. Our Domain Rating is about 42. Our voice is calm, practical, and reassuring, never twee. We target US English. House style for titles is under 60 characters, sentence case, no clickbait. Always flag the search intent (informational, commercial, or transactional) for any keyword.
Web search reads the live SERP so you target what you can win, not what looks big.
Search the live SERP for 'monstera leaves curling' and tell me: the search intent, the top 5 ranking pages and their angle, the content format Google is rewarding (listicle, how-to, video), the rough word count of the top results, and the People Also Ask questions. Then say whether Bloomfield can realistically rank given our DR of 42.
Give me 15 related long-tail keywords around monstera care problems, each with the likely intent and a one-line note on whether it's a quick win or a long play for a DR 42 site. Group them by subtopic.
Of those 15, which 5 would you publish first and why? Order them by effort-to-payoff.
What you get: A ranked shortlist of keywords you can actually rank for, sized against your real domain strength.
One keyword becomes a full brief, then a saved Skill you reuse on the next one.
Build a content brief for the article targeting 'why are my monstera leaves curling'. Include the primary keyword and 5 secondary keywords, the search intent, a recommended H1 and URL slug, a full H2 and H3 outline that covers what the top results cover plus one angle they miss, the People Also Ask questions to answer, a target word count, and 3 internal links we should include. Keep our calm, practical voice.
Add a short 'what good looks like' note for the writer: tone, what to avoid, and the one thing this piece has to do better than the current top result.
Save this as a Skill called 'Content brief' so I can generate one for any keyword by pasting it.
What you get: A complete brief a writer can run with, and a saved Skill that turns the next keyword into a brief in one paste.
Plan the pillar and the supporting pieces so the work compounds instead of scattering.
I want to own 'monstera care' as a topic. Design a content cluster: one pillar page and the supporting articles around it. For each, give the working title, the target keyword, the intent, and how it links back to the pillar. Aim for 8 to 12 supporting pieces and flag any that overlap or risk cannibalizing each other.
Now lay the cluster out as a table with columns for title, keyword, intent, funnel stage, and priority, so I can drop it straight into our planning doc.
What you get: A mapped cluster with internal links planned up front, so the topic builds authority instead of competing with itself.
Batch the metadata work the calendar always demands and nobody enjoys.
Here are 12 article titles and their primary keyword. For each, write 3 title-tag options under 60 characters in sentence case and one meta description under 155 characters, in our voice, no clickbait, keyword near the front. [paste the list]
For the 'monstera leaves curling' piece, give me 5 title variations testing different angles: problem-first, question, how-to, reassurance, and number-led. Tell me which you'd ship for a worried beginner and why.
What you get: Clean, on-brand titles and metas for a whole batch of pages in one pass.
This is the Keimodel angle: same brief, different model, side by side, so you pick the stronger opener.
Write the opening 120 words for the 'why are my monstera leaves curling' article: lead with the reassurance that it's usually fixable, name the three most common causes, and set up the rest of the piece. Our calm, practical voice, no fluffy preamble.
That was on the current model. Now switch the underlying model to GPT and write the same intro from the same brief, then put the two side by side and tell me which opens stronger for a worried beginner and why.
Take the best lines from each and write one final 120-word intro.
What you get: A real read on which model's draft fits your voice, so you choose the writing tool on evidence instead of habit.
Upload the sitemap and let the agent do the linking nobody has time to map by hand.
I'm uploading our blog sitemap as a CSV of URL and title. For the new 'monstera leaves curling' article, suggest the 6 most relevant existing posts to link to, with the exact anchor text to use for each and a one-line reason it's a good contextual match.
Now go the other way: from the same list, which 5 existing articles should add a link to this new piece once it's live, and what anchor text would make sense on each?
What you get: A two-way internal-linking plan with real anchor text, built from your actual site instead of guesswork.
Scheduled tasks keep the SERP and your Search Console in view without you remembering to look.
Every Monday at 8am, search for what's newly ranking in the houseplant-care space: new questions trending, fresh People Also Ask, and any competitor articles that have started ranking for our target keywords. Send me a 6-bullet digest with one suggested article to write.
On the 1st of each month, remind me to upload the latest Search Console export so we can find pages stuck on page two that are close to breaking through.
What you get: A weekly digest of fresh ranking opportunities and a monthly nudge to rescue the pages sitting just off page one.
Content marketing
Draft long-form, atomize it everywhere, and compare models to pick the best line.
Marketing agencies
Briefs, calendars, ad variations, and client reporting at agency speed.
E-commerce / DTC owners
Pricing math, product copy in your voice, competitor research, and customer replies.
Open the Agent, paste any prompt above, and change the details to fit your business.