The Lumina SEO blog: expert analysis on search engine optimization, generative engine optimization, and the evolving AI search landscape.
What AI visibility means, the four metrics worth tracking, and why a rank number is the wrong answer. Live audit across ChatGPT, Claude, Gemini and Perplexity: turning on web search costs up to 24 times an ungrounded call, Gemini never names the publisher it cites, and in German AI Overviews YouTube outranks every SEO vendor with 3,118 citations.
All 5 sitemap variants, the four submission methods (one deprecated in June 2023), the lastmod field everyone misuses, and IndexNow — the push protocol every guide in the SERP missed. Live audit of 10 top sitemap guides: 0 of 10 mention IndexNow, 0 of 10 warn about lastmod inflation, and Semrush DE's #2 result hasn't been touched since June 2021.
How image SEO actually works in 2026: modern formats, alt text, LCP priority hints, schema, and the AI-search layer. Live audit of 10 top guides for "image seo" — 0 of 97 images use WebP or AVIF, 1 of 10 wires author sameAs, and the rank-4 google.de article hasn't been touched in 1,411 days while its title still says "2026".
How E-E-A-T actually works in 2026, why Author Entity verification is now an AI-citation signal, and the schema patterns that get you cited. Live audit of 10 top-ranking guides for "e-e-a-t seo" — only 4 ship FAQPage, OMT at google.de rank 2 hasn't been updated in 754 days, and Marktgetrieben ranks #4 with zero JSON-LD schemas.
What robots.txt actually does, the 19 AI crawler user-agents you have to know in 2026, the five patterns that work, the six mistakes most sites make, and the audit method that catches all of them. Live audit of 10 top-ranking guides on Google EN + DE — only 1 ships FAQPage, 3 don't even mention AI crawlers, and the contrarian take from Heise's "robots.txt obituary" is wrong.
How internal linking actually shapes rankings, the five patterns that work, the six mistakes most sites make, and the AI-search angle nobody else writes about. Live audit of 11 top-ranking guides on Google EN + DE — only 2 ship FAQPage, none cover how AI crawlers read internal links, one DE article has been stale for 1,374 days, and the rank-1 DE result ships zero Article schema.
How keyword cannibalization actually drags rankings, when it's a real problem versus a non-issue, the three diagnostic methods that find it cleanly, and the six fixes ranked by impact. Live audit of 10 top guides — only 1 ships FAQPage, none cover how AI search engines handle the same question, and one ranks #4 on Google.de with zero structured data at all.
How Google actually renders JavaScript pages, why client-side rendering still breaks SEO in 2026, and the new constraint nobody else is talking about: AI crawlers do not run JavaScript at all. Live audit of 12 top JS SEO guides on Google EN + DE — zero ship FAQPage, four still teach Dynamic Rendering despite Google's own docs calling it “a workaround”, the most-cited guide is 2,245 days stale.
AI text doesn't fail because GPTZero catches it. It fails because the patterns it ships with quietly drop engagement signals. The 8 patterns to fix, the German-specific ones from KONVENS 2024, and a live audit of 9 articles and tools currently ranking on the topic — 0 of 9 ship FAQPage, 5 of 5 DE results are tool homepages, no editorial competitor in the DE SERP at all.
The 12 directories worth building by hand, the data aggregators that distribute your NAP for free, and the bulk-listing services worth nothing in 2026. Live audit of 9 top-ranking guides: 8 of 9 ship no FAQPage, average dateModified staleness 557 days, 5 of 9 still recommend Foursquare with no nuance about the 2024 consumer-app shutdown.
Every field that moves a ranking, the verification flow per category, the suspension rules nobody documents. Live audit of 10 top-ranking GBP guides: 4 of 10 ship no FAQPage, 5 of 10 are silent on freshness, 0 of 10 use @id refs for both author + publisher, and the #1 EN result blocks crawlers entirely with a Cloudflare challenge.
The three factors Google actually ranks by, the review numbers that separate the 3-pack from position 7, and the myths that waste your budget. Live audit of 10 top-ranking Map Pack articles: 5 of 10 ship no FAQPage, 3 of 10 have no dateModified, OMR ships a 415-day-stale guide on page 1.
LCP, INP, CLS — what they measure, what counts as good, and why your scores disagree between PSI and Search Console. Live audit of 10 top-ranking CWV articles: 0 of 5 EN pages pass Core Web Vitals themselves. Ahrefs: PSI 81 lab, INP 390 ms field.
The types, the code, the validators, the mistakes. Live audit of the 5 top-ranking schema-markup guides on Google — 0 ship FAQPage, 380-day avg dateModified staleness, Umbraco covers 9,563 words on schema with only a bare WebPage block.
PSI Mobile + Desktop against the 50 largest DACH news sites. Median mobile score: 44/100. Median LCP: 8.9 s. Only puls24.com passes. br.de hit a 33-second LCP. Tech stack detection included.
Full reference for 12 major AI bots — GPTBot, ClaudeBot, Perplexity and friends. Live audit of 10 top news sites: 50% block ClaudeBot, only 10% block Claude-SearchBot. The training-vs-retrieval split works.
Live audit of 10 top-ranking articles across US and DE markets. 0 of 10 ship FAQPage schema. 164-day average dateModified staleness. What schema actually gets cited by ChatGPT and Perplexity.
50 DACH media tested for JS rendering. Seven hide 25 %+ of their content from AI crawlers. News.at shows 100 % only after JS runs — classic AI crawlers see a blank.
I ran the robots.txt of 50 of the largest media sites in DACH through a full AI crawler check. 52% block GPTBot. 0% block Googlebot. Full data and findings.
Three optimization disciplines, one web. SEO chases blue-link rankings. AEO wins direct-answer queries. GEO gets cited inside generative AI. Here is how they differ and how they stack.
Practical SEO for ChatGPT, Perplexity, Gemini, and AI Overviews. Six tactics that move the needle, with live audit data on the articles currently ranking — and where they miss.
GEO optimizes content to be cited inside ChatGPT, Perplexity, and Google AI Overviews. SEO optimizes for blue links. Here is the honest breakdown of what changed and where each one wins.
Google updated its documented crawl limit from 15 MB to 2 MB for HTML pages. Here's what changed, who's actually affected, and how to check your own pages.
The March 2026 Core Update introduces Gemini 4.0 Semantic Filter, stricter E-E-A-T requirements, and new INP thresholds. An expert analysis of the impact and strategic recommendations.
Google just launched a new user-agent for AI-powered web interactions. Here's what Google-Agent means for your website and how to prepare.
Lumina's SEO blog covers the stuff I actually work on day to day: Google algorithm updates, the shift from classical SEO to GEO (generative engine optimization), AI crawler behavior, and the practical tools you need to keep up. No repackaged press releases or recycled tips. Every post is based on my own research or something I tested myself before writing about it.
My filter is simple. If I had to explain something to a client and couldn't find a clear write-up to send them, that's a post. Most articles here started as an email I drafted twice.
For the same content, the format makes a big difference. Blogs are the best fit for technical topics that need code snippets, diagrams, links, and reference tables you can copy. Newsletters win when the value is an opinion or a curated list delivered regularly. Podcasts are best for interviews and the kind of context you can't compress into bullet points. I stick with the blog format here because most SEO topics need exact syntax and screenshots to be useful.