GreenSip is a fictional brand created for the capstone of SEO Mastery: From Fundamentals to GenAI and GEO Strategies on Coursera. The author, publisher and URLs used in the schema were provided by the course. Since the page was never published, there is no traffic data. The results below are the deliverables and their quality checks.
The challenge
GreenSip is a wellness tea brand built on honesty: no miracle detox, no promises of rapid weight loss. It needed a page for one of the highest-demand questions in its niche, “green tea for weight loss”, with 33,100 monthly searches in the US.
That creates a tension. The query attracts hype-heavy content, but weight loss is a YMYL (Your Money or Your Life) topic, where Google expects accuracy and caution. The page had to rank for a popular query without making the claims that usually come with it, and it had to be structured so AI systems could understand and cite it.
The strategy
The project combined three layers, each with a clear role:
- Traditional SEO as the foundation: one keyword cluster mapped to one page, with every on-page element built around it.
- GEO thinking for the structure: clear entities (green tea, catechins, EGCG, caffeine, matcha, extract supplements) and a clear outcome for the reader: an informed, safe choice, not a promise.
- AI for speed, with human judgment on every output: AI generated options and drafts, and I selected, fact-checked and rewrote them.
The dominant intent was set as informational, with a layer of commercial investigation. The page’s job is to reach people at the top of the funnel with useful, honest answers, and then guide them toward the brand.
Execution
1. Keyword research and clustering
I expanded the topic with WordStream’s keyword tool (US market) and selected one primary keyword plus five supporting terms with the same search purpose. Lower-volume terms were kept on purpose: they usually mean less competition and add depth to the page.
| Keyword | Intent | Monthly searches (US) |
|---|---|---|
| Green tea for weight loss (primary) | Informational | 33,100 |
| Tea to drink to lose weight | Informational | 9,900 |
| Best tea to lose weight | Commercial investigation | 9,900 |
| Beneficial teas for weight loss | Commercial investigation | 6,600 |
| Matcha for weight loss | Informational | 6,600 |
| When to drink green tea for weight loss | Informational | 480 |
2. On-page plan
All six terms were mapped to a single page at /green-tea-for-weight-loss/. The plan defined the title tag, H1, a full H2/H3 outline, the meta description, image alt text, internal links to related pages (green tea vs. black tea, matcha benefits, the product page) and FAQs written in the users’ own words.
3. AI-assisted refinement
With the plan ready, I used generative AI to produce several options for each element, combining role-based and few-shot prompting, then chain-of-thought prompting for the GEO and AEO improvements. I picked the strongest option for each element based on clarity, keyword alignment and brand fit.
| Element | Before | After | Why it’s better |
|---|---|---|---|
| Title tag | Green tea for weight loss: what the science says │ GreenSip | Green tea for weight loss: realistic benefits and safe use | The keyword leads, it stays under 60 characters, and it sets realistic expectations instead of hype |
| Meta description | Does green tea really help with weight loss? See what the science says, when to drink it, and how it compares to matcha and other teas. | Does green tea help with weight loss? See what studies show, how big the effect really is, and how to fit it into a balanced routine. | Mirrors the searched question, keeps one clear promise and ends with a call to action |
| H1 | Green tea for weight loss: does it really work? | Green tea for weight loss: how it helps and how to use it well | Drops the yes/no framing and promises both an answer and practical guidance |
AI also produced the first draft of the article. I reviewed it by hand for factual accuracy, natural keyword use, readability and GreenSip’s warm, evidence-based tone.
4. Answer-ready structure
To make the content easy for AI systems to parse and cite, I added three elements:
- An answer capsule of 40 to 60 words at the top of the article, giving the complete answer in one paragraph.
- Question-style H2s, such as “Does green tea help you lose weight?” and “Which mistakes cancel the benefits?”
- Six self-contained FAQs, each opening with a direct answer followed by specific, cautious details on dosage, timing, results and safety.
The answer capsule:
Green tea may modestly support weight loss through its catechins, mainly EGCG, and caffeine, which slightly raise calorie and fat burning. Most studies used 2–4 cups a day, each with about 20–45 mg of caffeine. The effect is small and most useful alongside diet and exercise; pregnant people and caffeine-sensitive drinkers should limit intake.
5. Structured data
I built two JSON-LD blocks: BlogPosting (with author, publisher, logo and a stable @id) and FAQPage (matching the six FAQs word for word). Both were created with the TechnicalSEO.com Schema Markup Generator, without AI, and checked against the visible page content to avoid mismatches.
Results
- A 6-keyword cluster covering 66,500+ combined monthly searches in the US, mapped to a single page with clear intent.
- A complete on-page plan: title, meta description, H1 to H3 outline, alt text, internal links and FAQs.
- Documented before and after improvements for every key on-page element.
- An answer-ready article with an answer capsule, question-style headings and six self-contained FAQs.
- BlogPosting and FAQPage schema validated with zero errors on the Schema.org Validator.
If the page went live, I would track organic traffic and rankings for the cluster, visibility in AI-generated answers, click-through rate from search, and time on page, and update the content as research and search behavior change.
What I learned
- On a YMYL topic, honesty is the strategy. Setting realistic expectations and adding a “who should be careful” section makes the page more trustworthy for readers and search engines alike.
- AI is fast at options, and the real work is choosing. Generating three titles takes seconds. Knowing which one fits the intent, the brand and the facts is the part that matters.
- Schema has to mirror the page. Structured data only helps when it matches what users actually see, so every FAQ in the markup is identical to the one on the page.