SEO & GEOParis 2026

GEO is the art of optimizing a site to be cited by AI answer engines — ChatGPT, Claude, Perplexity, Gemini, Google AI Overviews. The term comes from a Princeton paper (arXiv 2311.09735, 2023). Concretely: structured data, llms.txt, answer-capsules, E-E-A-T authority, well-sourced factual content. Classic SEO optimizes for 10 blue links. GEO optimizes to be the sentence the AI quotes in its answer.

Not exactly — they're transforming it. 48% of SERPs have an AI Overview in 2026 (BrightEdge), and CTR at position #1 drops 34.5% when one appears (Semrush). But 86.5% of top-ranking pages now use AI in their content (Ahrefs), and pages cited in AI Overviews capture very high-intent qualified traffic. The real answer: SEO + GEO, not one against the other.

Technical quick wins (INP, indexation, schema): 2-4 weeks. AI citations on niche queries: 4-8 weeks. First Google ranking gains on long-tail: 2-3 months. Top-3 dominance on competitive queries: 6-12 months. SEO/GEO is a long-term investment with average ROI of 748% over 3 years in B2B (SEOProfy 2026).

Yes, but with serious cleanup. Default WordPress often fails INP (43% of sites do). You need: a lightweight or rebuilt theme, caching (WP Rocket), a CDN, AVIF images, plugin cleanup, and often page-builder code replaced with clean code. If technical debt is too heavy, I recommend Next.js. Otherwise, optimized WordPress remains viable — I've already taken a WordPress construction site from page 5 to top 3 in its Île-de-France sector.

llms.txt is a markdown file proposed by Jeremy Howard (Answer.AI, September 2024) that summarizes your site for LLMs: key pages, answer capsules, context. It's the equivalent of robots.txt but for generative models. Status in 2026: 800,000+ sites have published one — Anthropic, Cloudflare, Stripe, Vercel, Supabase among them — but no major LLM officially parses it yet (Google confirmed this). Why do it anyway? It's a one-hour job, it cleanly structures your content for the day adoption catches up, and several GEO tools already use it as a signal.

Core Web Vitals have been a ranking factor since 2021. INP (Interaction to Next Paint) replaced FID in March 2024, and 43% of sites fail it. Bad INP = Google demotion + bounce rate spike. 2026 thresholds: INP < 200ms, LCP < 2.5s, CLS < 0.1. I measure via PageSpeed + CrUX and fix at the stack level (lazy images, code splitting, third-party scripts, React hydration).

A 2020 SEO agency sells you backlinks, PowerPoint audits and 30 Yoast-optimized articles a month. Me: real technical audit (Screaming Frog, CrUX, Lighthouse), GEO included from day 1 (llms.txt, LLM citations, AI Overviews), honest tracking (GSC + GA4 + Semrush + LLM monitoring). No junior team learning your project, no cosmetic reporting. You speak to the person who executes.

Google Ads = immediate traffic that stops the day you cut budget. SEO/GEO = traffic that compounds over time. Average ROI 748% over 3 years in B2B (SEOProfy 2026), and the marginal cost of a visitor tends toward zero. My advice: Ads to quickly test which angles convert, SEO/GEO to build a long-term asset. The two are complementary, not competing.

2026 stack: Google Search Console + GA4 + Tag Manager for the basics; Semrush for competitive intelligence; Hotjar to understand real behavior; Screaming Frog for technical audits; PageSpeed Insights + CrUX for Core Web Vitals; and a custom monitoring of ChatGPT, Claude, Perplexity, Gemini citations on your target queries. All consolidated in a monthly dashboard.