I used Claude to rank on Google in 9 languages. Here’s the exact system

No content team. No agency. One AI, nine markets, and a process that takes about an hour per article.

When I started the MaviBot blog, the strategy seemed obvious. Find high-volume keywords, write articles, get traffic.

Then I checked the search results.

"WhatsApp chatbot": first result, G2. Second, Capterra. Third, HubSpot. Sites with DR 80+, thousands of backlinks, years in the index. A new domain doesn't belong there. That's not pessimism, that's just how SEO math works.

So I tried something different.

First, pick the market. Then find the keywords.

MaviBot operates in 9 language markets: Arabic, English, Spanish, French, Kazakh, Portuguese, Russian, Turkish, Ukrainian. Each language has its own blog, its own Google index, its own level of competition.

The same keyword that's impossible to rank for in English can be wide open in Portuguese.

Brazil is the second-largest WhatsApp market in the world. Restaurants there literally take orders through WhatsApp. I searched "chatbot para WhatsApp restaurante" and found almost nothing: a few forums, a couple of old articles, no real SaaS product with a proper piece of content targeting that query.

High demand. No competition. That's the only signal you need.

That became the whole approach: don't fight for positions that are already taken. Find markets where there's room.

How Claude turned this into a system
Identifying the market is strategy. Executing it at scale is a different problem.

Writing one well-optimized article manually takes several hours. Writing that article for nine language markets means hiring a full content team.

I use Claude with SEO skills through Claude Code: a structured pipeline with specific steps, not a one-off prompt.

Step 1: Topic and angle research

The searchfit-seo:create-topic skill analyzes the query: it evaluates competitiveness, suggests the content angle, and determines what format will actually rank. For the WhatsApp restaurant bot query, it landed on a HowTo format: a step-by-step guide with concrete actions.

Step 2: Writing the article

The searchfit-seo:create-content skill writes the full article with a proper H1→H2→H3 hierarchy, short paragraphs, and specific examples. Not the generic "chatbots are useful for business" angle, but something built for a specific market and a specific query.

Step 3: The technical SEO wrapper

This is what most writers skip.

Every article comes out with a complete frontmatter block:

title: "WhatsApp Chatbot for Restaurants: Take Orders Automatically"
description: "Use a WhatsApp chatbot to take orders, confirm bookings, and
  recover abandoned carts automatically. Run your restaurant on WhatsApp
  with MaviBot. Start free."
slug: "whatsapp-chatbot-restaurant-food-delivery"
canonical: "https://mavibot.ai/posts/whatsapp-chatbot-restaurant-food-delivery"
keywords: ["WhatsApp chatbot restaurant", "WhatsApp food ordering bot",
  "restaurant chatbot WhatsApp"]

The title is capped at 60 characters so it doesn't get cut in search results. The meta description stays under 155 characters and includes a CTA. The canonical URL prevents duplicate content issues across translations.

Step 4: Schema markup

This is what gets you the rich snippets in Google, those cards with steps or questions that appear directly in the results. Claude generates them automatically from the article structure.

For step-by-step guides, HowTo schema:

{
  "@type": "HowTo",
  "name": "How to Set Up a WhatsApp Chatbot for Your Restaurant",
  "totalTime": "PT60M",
  "estimatedCost": {
    "@type": "MonetaryAmount",
    "currency": "USD",
    "value": "0"
  },
  "step": [
    {
      "position": 1,
      "name": "Connect Your WhatsApp Business Account",
      "text": "In MaviBot, go to Channels and connect your WhatsApp Business
        number via the API. Takes about 10 minutes."
    },
    {
      "position": 2,
      "name": "Upload Your Menu to the AI Knowledge Base",
      "text": "Add your menu — dishes, prices, allergens, delivery details.
        Set reminder rules for abandoned orders."
    }
  ]
}

FAQPage schema gets added separately for the Q&A section. Both blocks ship in the same file as the article.

Step 5: Market-specific instructions for translators

Each article includes a comment block with a breakdown per market:

🇧🇷 Português — Brazil (Mercado Pago, Pix, massive food delivery market)
🇪🇸 Español — Mexico (OXXO), Colombia (PSE), Argentina
🇸🇦 العربية — UAE, Saudi Arabia, Egypt (PayTabs)
🇹🇷 Türkçe — Turkey (largest food delivery market in the region)

The translator doesn't just get "translate this." They get specific instructions about which local payment methods to mention, which market nuances to emphasize, which details to localize.

Step 6: Internal links

Claude maps internal links across articles by topic: the restaurant bot article links to the WhatsApp small business automation piece, which links to the broadcast campaigns guide. That's both user experience and a signal to Google about site structure.

What one cycle produces
A single session, from topic selection to finished file, takes about an hour.

That file goes to translation. The searchfit-seo:translate-content skill adapts keywords for local search rather than just translating text. "Chatbot para WhatsApp restaurante" for Brazil and "chatbot WhatsApp restaurante" for Spain are different queries with different search volumes. The skill handles that distinction.

One article. Nine languages. Nine chances to rank in markets where the competition is thin.

What Claude doesn't replace
The system only works if the topic and market are chosen correctly to begin with.

Claude doesn't know that the Brazilian WhatsApp automation market is underserved. That comes from checking the search results yourself, understanding where your product actually operates, and making a judgment call. Strategy stays with the human.

It also doesn't replace editing for brand voice. The draft that comes out of the pipeline is a starting point, not a final text. I go through it manually: cut generic phrasing, add specifics from real product experience, make it sound like us and not like every other SaaS blog.

And one more thing: fresh data. Claude doesn't know what happened in the market last week. Stats and current case studies get added by hand.

Why this works better than hiring a content team
Not because it's cheaper. Because it scales.

A copywriter writes well, but they're one person. They don't carry technical SEO requirements in their head, won't generate schema markup for every article, and won't translate a piece into nine languages with localized keyword adaptation.

The combination (strategy from the team, production through AI, editing for brand voice) covers everything without building a full department.

We didn't automate content. We automated the routine inside the content process. Those are different things.

If you run a SaaS blog and haven't tried this approach, start with one market where there's room. Sometimes an empty search results page is the best thing you can find.