By Khalid Namra15 min read
What exactly happened?
For two decades the results page was a ranked list and the user's job was to choose. Today a large part of the screen is a written answer generated by a language model from several sources at once, with a few small citation links beside it. Many searches now end with no click at all, because the user already got the answer.
The key idea: ranking is not gone, but it is no longer the finish line. The goal is to be one of the sources the answer is built from.
Three different layers, not one
| Layer | Where | What governs it |
|---|---|---|
| Google AI Overviews | Top of the Google results page | Google's own index: if you are not well indexed, you are not cited |
| Conversational assistants | ChatGPT, Gemini, Claude | A mix of training knowledge and live retrieval through partner indexes |
| Answer engines | Perplexity and similar | Always-live retrieval with explicit source citation |
The common mistake is treating all three as one. Overviews depend directly on your Google strength, assistants depend on how widely your brand is mentioned across the web, and answer engines depend on how quotable your page is at question time.
What this really does to your traffic
- Short informational queries lose clicks because the answer is displayed in full.
- Complex, comparison and commercial queries keep their value because the user needs detail, price and contact.
- The quality of arriving visits rises: someone who clicks after reading a summary is closer to deciding.
- A brand mention inside the answer has value even without a click; it builds familiarity and trust.
Distrust any absolute figure about lost traffic. Impact varies drastically by query type, sector and market. Measure it on your own Search Console data, before and after.
The situation in Arabic content
Arabic web content is far smaller and weaker than English, and much of it is poor machine translation. Models therefore find fewer trustworthy Arabic sources and fall back on translating English ones or leaning on a handful of sites. That creates a rare opportunity: an original, accurate, well-structured Arabic page in a specialised topic can become the preferred source far faster than in English.
What to do now: five actions
- Fix the technical base: speed, crawling, indexing. A model cannot cite what it cannot read.
- Rebuild key pages as question, then a direct two-line answer, then detail.
- Define your entity clearly: an about page, Organization data, one consistent name, consistent mentions across platforms.
- Add what a model cannot generate: your market numbers, local examples, first-hand experience, original tables.
- Build a fixed prompt set and test it monthly across engines so you know your position by data, not impression.
Frequently asked questions
Is SEO dead because of AI?
No. Models rely on the same web index, and a site that cannot be crawled or indexed appears in neither. What changed is the final output and how you measure it, not the foundation.
Should I block AI crawlers?
It is a commercial decision. Publishers selling content may block. Companies that want customers should not: blocking means disappearing from the answers their customers read.
How do I know if my brand is cited in answers?
There is no official report. The practical method is testing a fixed prompt set on each engine periodically and logging whether you appeared and alongside whom — which is what the SADARA platform automates.
Related SADARA services
- AI Search visibilityUsers no longer read ten results, they read one answer. Our job is to place your brand inside that answer.
- GEO — Generative Engine OptimizationGEO is not SEO renamed. It is a different way of structuring knowledge so models can understand and cite it confidently.