Over the last few months I reviewed the SEO of all my websites, and my clients' sites, with Claude Code (Opus 5.5) working alongside me. Not a chatbot handing out generic tips: an AI that reads the site's code, the Search Console and Analytics data, and checks every fix on the live site. The result: problems that had been invisible for months. Here is the method and the real cases.
1. Actually reading the data: Search Console and Analytics
I start from Google Search Console reports: queries, pages, coverage, 404 errors, pages "indexed, though blocked by robots.txt". Claude cross-references that data with the site's code to find the cause, not just the symptom. Some findings on my own site:
- Visits were coming in, but few enquiries: the Spanish homepage talked about ".NET" while people search for "websites in Pucón". I refocused it on Pucón and the Araucanía region.
- The "made by" credits on my client sites all pointed to my homepage with the same text. They now point to specific service pages, with varied anchor text.
- The FAQ structured data had been invalid for a month because of one missing brace: Google silently ignored it.
- The blog generator copied the same meta description onto 17 articles.
- URLs ending in
.htmlor/answered as duplicate pages instead of redirecting.
2. Google Analytics 4, properly set up
No measurement, no decisions. I install GA4 with key events: WhatsApp clicks and quote requests. That way you know which page brings contacts, not just visits.
The data also needs cleaning: on one of my sites, the bot detector didn't recognise Googlebot, so its visits were counted as human. The statistics were inflated.
3. Bing Webmaster Tools and IndexNow: the forgotten channel
A GA4 figure from August 2026 on findyourneurotype.com: Bing and the AI assistants built on its index brought about six times more traffic than Google. Hardly anyone works on it.
- Bing Webmaster Tools: import from Search Console in two clicks, verification and sitemap included.
- IndexNow: a script notifies Bing as soon as a page is published. It gets indexed in minutes instead of waiting for the crawler.
4. Opening the site to AI engines
More and more people ask ChatGPT, Perplexity, Claude or Gemini instead of searching Google. To be cited, they need to be able to read you:
- A robots.txt that explicitly allows AI crawlers (GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot, Google-Extended…) and blocks SEO scrapers that only eat resources.
- An llms.txt: a plain-text summary for language models, with what you sell, prices, languages and key links.
- Schema.org structured data: person, local business, services, FAQ, breadcrumbs. AI understands a site that describes itself.
5. Real cases
FindYourNeurotype
A 1,277-URL sitemap in 6 languages, Bing Webmaster Tools and IndexNow. Search Console flagged some demo PDFs as "indexed, though blocked by robots.txt": the block was removed and a noindex header is used instead, which Google can only read if it's allowed to fetch the file. Case study (in Spanish).
ExploreYourDNA
A robots.txt open to AI engines and closed to scrapers, and a sitemap regenerated automatically every day. Case study (in Spanish).
FuelYourDNA
An llms.txt summarising the reports, prices and languages for AI assistants, and a 2,179-URL sitemap in 6 languages.
Parcelas Patagonia
The region pages were almost empty for months: the code filtered listings marked "Aprobado" while the database stored them as "Approved". The English blog content existed in the database but was never displayed. And the .cl domain had only its homepage indexed after 12 months: after moving to .com with 301 redirects, Google indexed the first pages the very next day. 34 city landing pages were added.
jeromestephan.com
Translated URLs in English and French, clean 301 redirects, 79 URLs crawled with zero errors, and a Spanish homepage refocused on Pucón and the Araucanía.
6. A blog guide that answers real searches
A blog isn't there just to "have a blog". Each article answers a specific question your customers type into Google (prices, comparisons, real cases), links to the matching service page and ends with a call to action: a quote or WhatsApp. With Claude I build the article plan from real Search Console queries.
What AI doesn't do on its own
Claude speeds up diagnosis enormously, but every fix gets verified: a live crawl of the site, structured data validation, redirects tested one by one. And positioning decisions (who to target, what to offer, in which area) remain human.
My offer: a site ready for Google, Bing and AI
- Technical audit of the code and of Search Console.
- GA4 with key events (WhatsApp, forms, quotes).
- Bing Webmaster Tools and IndexNow.
- robots.txt and llms.txt for AI engines.
- Schema.org structured data.
- An article plan based on real searches.
More details on my SEO in Chile page, or get in touch for a free audit.
Want your business to show up on Google and in AI answers?
I'll prepare a free audit of your website and Google profile, no strings attached.
Get my free auditMessage me on WhatsApp