AI Agent SEO Workflow: Automated Weekly Keyword-Gap Blogging

1 hour ago

8 min read

How HomeoAI gets four to five new doctors a day with no paid marketing, and how to build the same human-approved content engine for your product

Automated SEO blogging with AI agents: a weekly workflow that finds keyword gaps, drafts posts and publishes after human approval, by Codersarts.

Most small products have the same marketing problem. People are already searching for what you offer, but they can't find you because you have no page that answers their search. Fixing that by hand means keyword research, briefs, drafts and publishing every week, which a busy founder rarely keeps up.

Tejesh Vaish solved this for HomeoAI, the AI assistant he built for homeopathic practitioners. On the Codersarts Podcast, he explained his weekly AI agent workflow: it finds keywords users search for that HomeoAI doesn't cover yet, writes a blog post for each gap, sends it to him on WhatsApp for review, and publishes it once he approves. He does no paid marketing. HomeoAI still gets four to five new doctors every day from around the world, and it ranks first for its brand searches ahead of around ten copycat products.

This post breaks down that workflow and shows how to build your own, with the guardrails that keep automated content helpful rather than spammy.

Watch the episode: [ADD HOMEOAI EPISODE YOUTUBE LINK]

Why keyword-gap content works

A keyword gap is a search your audience makes that your site doesn't answer well. There are three common kinds:

Impressions without clicks: Google already shows your site for a query, but on page two or lower, because no page targets it directly.
Competitor keywords: searches your competitors rank for and you don't.
Unanswered questions: questions your users ask in support chats, sales calls or the product itself.

Tejesh described the logic plainly: people are searching for these keywords but not finding your posts, so if you create posts around those keywords, people start finding you. Each gap you close is a small, permanent source of the right visitors.

How HomeoAI's workflow runs

Step	What happens	Who does it
1. Run weekly	The workflow starts every seven days	AI agent, on a schedule
2. Find gaps	Compares search data from Google Search Console and SEO tools with existing content to find keywords users search for but HomeoAI doesn't cover	AI agent
3. Draft	Writes a blog post for each gap	AI agent
4. Review	Sends the draft to Tejesh on WhatsApp	Agent sends, Tejesh reads
5. Approve	Tejesh replies to approve or request changes	Tejesh
6. Publish	The agent deploys the approved post	AI agent

Two choices make it work. It runs on a schedule, so content keeps shipping even in busy weeks. And it has a human approval step, so nothing reaches the site without a person checking it.

He also built comparison pages, such as HomeoAI compared with ChatGPT, Gemini, Claude and Perplexity. People comparing tools are close to choosing one, so these pages bring in high-intent visitors.

Build your own: the 7-step workflow

Step 1: Connect your search data

Start with the data you already own:

Google Search Console: queries, impressions, clicks and average position for your site.
Google Analytics: which pages bring visitors who sign up or buy.
An SEO tool such as Semrush, for competitor keywords and search volumes.
Your own questions: support tickets, sales notes and in-product searches.

Tejesh also mentioned using an open-source SEO tool connected to Google Analytics and Search Console to audit landing pages and suggest fixes.

Step 2: Find the gaps

Filter for queries where:

Impressions are meaningful but position is low (for example, below position 10).
No existing page clearly targets the query.
Competitors rank and you don't.

Step 3: Prioritise

Not every gap deserves a post. Score each one on:

Business relevance: would someone searching this become a user or customer?
Intent: comparison and problem searches usually convert better than broad definitions.
Effort: can you write the best answer on the web with what you know?

Pick the top one to three each week.

Step 4: Draft with an AI agent

Give the agent what it needs to write accurately:

The keyword and the search intent behind it.
Your product facts: features, pricing, limits and real examples.
Your style guide: tone, structure and words to avoid.
Internal links to existing pages it should connect to.

A brief template you can adapt:



Write a blog post targeting the search query: "{keyword}".
Reader: {who searches this and what they want to decide or do}.
Use only these product facts: {facts}. Do not invent features, numbers or customers.
Structure: direct answer first, then sections that fully answer the query, then a short FAQ.
Link to these pages where relevant: {internal links}.
End with one clear next step: {call to action}.
Flag any claim you are unsure about as [CHECK].

Step 5: Human review

This is the step that keeps quality high. Send each draft somewhere you'll actually see it: WhatsApp for Tejesh, or Slack or email for a team. The reviewer checks:

Facts and claims about the product and the domain.
Usefulness: would a real reader get a complete answer?
Accuracy in sensitive domains: health, legal and financial content needs an expert's eye.
Brand fit: tone and positioning.

Reply to approve, edit or reject. A rejected draft costs nothing; a published wrong one costs trust.

Step 6: Publish and connect

After approval, the agent publishes the post with:

A title tag and meta description targeting the keyword.
Internal links to and from related pages.
Structured data such as FAQ schema where it fits.
A sitemap update, then a request for indexing in Search Console.

Step 7: Measure and refresh

Each week, the same workflow can check how earlier posts are doing: positions, clicks and sign-ups. Posts that rank on page two are often one update away from page one. Refreshing them is usually faster than writing new ones.

A simple gap-finder in Python

This sketch pulls queries from the Google Search Console API and flags likely gaps: queries with enough impressions, a low average position and no matching existing title. It assumes you have set up API credentials for your property.



from googleapiclient.discovery import build

def fetch_queries(credentials, site_url, start_date, end_date, row_limit=5000):
    service = build("searchconsole", "v1", credentials=credentials)
    body = {
        "startDate": start_date,
        "endDate": end_date,
        "dimensions": ["query"],
        "rowLimit": row_limit,
    }
    response = service.searchanalytics().query(siteUrl=site_url, body=body).execute()
    return response.get("rows", [])

def find_gaps(rows, existing_titles, min_impressions=50, min_position=10):
    titles = [t.lower() for t in existing_titles]
    gaps = []
    for row in rows:
        query = row["keys"][0]
        if row["impressions"] >= min_impressions and row["position"] > min_position:
            if not any(query.lower() in title for title in titles):
                gaps.append({"query": query, "impressions": row["impressions"],
                             "position": round(row["position"], 1)})
    return sorted(gaps, key=lambda g: g["impressions"], reverse=True)

# rows = fetch_queries(creds, "https://www.example.com/", "2026-10-01", "2026-10-31")
# gaps = find_gaps(rows, existing_titles=my_post_titles)
# Send the top gaps to your drafting agent, then to a person for approval.

Matching queries against titles is a rough first filter. In practice, also compare against page content and group similar queries into one topic, so you write one strong post instead of several thin ones.

Guardrails: helpful content, not spam

Automation makes it easy to publish too much, too fast. Google's position is that it rewards helpful content however it's produced, but its spam policies treat mass-producing low-value pages to manipulate rankings as spam. Keep your workflow on the right side of that line:

Publish less, better. One to three strong posts a week beats twenty thin ones.
Always keep the human approval step. It's what made HomeoAI's workflow safe.
Add what only you know: product details, real examples, data and expert judgement.
Don't invent facts. Instruct the agent to use only supplied facts and flag uncertain claims.
Avoid near-duplicates. Merge similar keywords into one comprehensive post.
Be careful with health, legal and financial topics. Have an expert review anything that could affect someone's decisions.

How we use this at Codersarts

We run scheduled AI tasks for our own marketing in a similar way: agents research topics that can improve our rankings, draft content for our services and our Medium publication, and surface promotion opportunities. Every piece is reviewed and published by a person. The agents do the repetitive research and first drafts; people make the decisions and keep the quality bar.

For a related look at how AI agents are changing engineering work, watch our Codersarts Podcast episode with Tejesh on AI coding agents and Claude Code:

AI SEO workflow checklist

Area	Question to ask	Good answer
Data	Where do keyword gaps come from?	Search Console, analytics, an SEO tool and real user questions
Gap rules	How is a gap defined?	Meaningful impressions, low position and no page targeting it
Priority	Which gaps get written first?	Highest business relevance and buyer intent, one to three a week
Inputs	Does the agent have real facts?	Product facts, style guide and internal links supplied in the brief
Accuracy	How are invented claims prevented?	Agent told to use only supplied facts and flag uncertain ones
Review	Who approves before publishing?	A named person, every time, through a channel they check daily
Publishing	Is each post fully set up?	Title tag, meta description, internal links, schema and indexing request
Quality	Are we avoiding thin or duplicate pages?	Similar keywords merged, fewer and stronger posts
Measurement	Do we know what's working?	Weekly tracking of positions, clicks and sign-ups, with refreshes

Build your AI content and growth workflows with Codersarts

HomeoAI's growth came from a simple system that runs every week with a person in control. Codersarts builds the same kind of engine for your product:

GTM engineering: automated workflows for keyword research, content drafting, approvals and publishing, connected to your search data and CMS. See https://www.codersarts.com/engineering/gtm.
AI agent development: scheduled agents with human approval steps, built on your tools and channels.
Technical content: expert-reviewed technical articles, case studies and comparison pages for developer and B2B audiences.
SEO and AI search visibility: audits, internal linking, structured data and content that AI assistants can cite.

Talk to us: https://www.codersarts.com/contact

FAQ

Can AI agents write SEO blog posts automatically?

Yes, AI agents can research keywords, draft posts and publish them on a schedule. The results are only as good as the facts you give them and the review step you keep. HomeoAI's workflow drafts automatically but publishes only after a person approves each post.

Does Google penalise AI-generated content?

Google says it rewards helpful content regardless of how it's produced. Its spam policies target mass-produced, low-value pages made to manipulate rankings. Fewer, accurate, genuinely useful posts with human review stay on the right side of that line.

What is a keyword gap?

A keyword gap is a search your audience makes that your site doesn't answer well. It shows up as queries with impressions but low rankings in Search Console, competitor keywords you don't rank for, or questions users ask that no page answers.

How often should an automated SEO workflow publish?

For most small products, one to three strong posts a week is enough. Consistency matters more than volume. Spend the rest of the time refreshing posts that already rank on page two.

What tools do I need for an AI SEO workflow?

Google Search Console and Google Analytics for your own data, an SEO tool for competitor keywords, an AI agent to analyse gaps and draft posts, a messaging channel for approvals, and access to your CMS for publishing.

Can Codersarts build an automated content workflow for our product?

Yes. Codersarts builds AI agent workflows that find keyword gaps, draft content, route it for approval and publish to your CMS, along with technical content and SEO support. Contact us at https://www.codersarts.com/contact.

Related reading:

Want to be a guest on the Codersarts Podcast? Email jitendra@codersarts.com with your name, role and the topic you'd like to talk about.