Is SEO dead with AI search?
No. Google’s AI Overviews and AI Mode pick their supporting links from pages that are indexed and eligible to show a snippet in Search, and Google says there are “no additional requirements” to appear in them (Google Search Central).
The overlap is measurable. Ahrefs analysed 1.9 million citations from 1 million AI Overviews and found that 76.1% of cited pages ranked in the top 10 organic results for the query (Ahrefs, updated 2 July 2026). That figure covers Google. ChatGPT and Perplexity run their own retrieval, so a top-10 ranking helps there less directly, but crawlable pages and clear topics still gate them.
What is the same in GEO and SEO?
The foundations: a crawler must be allowed in, the content must be in the HTML, the page must be indexed, and it must be about one clear topic. Generative engine optimization, GEO, adds nothing that replaces these, because an assistant cannot quote a page it never retrieved.
For a shop that means the SEO backlog still comes first: crawl rules, server-rendered product content, one product or one question per URL, internal links with descriptive anchors, and structured data that matches the visible text, which Google lists among its fundamentals for AI features.
What is different in GEO for an online shop?
Three things change. Answers quote passages rather than ranking whole pages, they lean on third-party sources more than search results do, and they reward evidence inside the passage.
AI search engines show “a systematic and overwhelming bias towards Earned media”, meaning third-party authoritative sources, over brand-owned content, unlike Google’s more balanced mix (Chen et al., arXiv, September 2025). In the foundational GEO study, adding cited sources, quotations and statistics raised a source’s visibility in generative answers by 30 to 40%, and by 115.1% for a source ranked fifth (Aggarwal et al., KDD 2024). Keyword stuffing went the other way: on GEO-bench it scored 17.8 against a 19.5 baseline on the paper’s position-adjusted word count.
The boundary is in the 2026 survey by Olivier Martinez, which reviewed 45 GEO studies and found no technique with a stable, cross-platform effect on discoverability or traffic, and that citation-aimed rewrites can reduce retrieval when they dilute the topic (arXiv 2607.14035). The GEO gains are real for passages already being read, and unproven beyond that.
| SEO | GEO | |
|---|---|---|
| Unit that competes | The page, for a ranking position | The passage, for a citation or a product mention |
| Main input | Query match, links, page quality | Retrieved passages, structured product data, third-party mentions |
| What helps | Relevance, internal links, earned links | The same, plus the answer in the first sentences, sourced numbers and quotations |
| What hurts | Thin or duplicate pages | The same, plus keyword stuffing and diluted topics |
| Outside sources | Backlinks as a ranking signal | Reviews, guides and press quoted directly in the answer |
| Success metric | Rankings, clicks, organic revenue | How often assistants name you, and orders arriving from agents |
| Measurement | Stable enough to read weekly | Varies from run to run, so repeated prompts and intervals |
Does Google need special files or markup for AI Overviews?
No. Google states that you do not need to create new machine-readable files, AI text files or markup to appear in AI Overviews or AI Mode. An llms.txt file is ignored by Google Search, and structured data is recommended for SEO rather than as an AI requirement.
Shopping agents and product feeds are a different case, because they compare offers on fields. The llms.txt guide separates what Google needs from what agents read.
What should a shop do first?
Fix the SEO foundations, then make each product and guide quotable, then work on the sources that name you. In that order, because each step depends on the one before.
- Crawl, render and index: product content in the HTML, crawlers allowed, one topic per URL.
- Product data as fields: price, stock, GTIN and delivery in JSON-LD and the product feed, through the Machine Readability Sprint.
- Answer first: each section opens with the answer in one or two sentences, with a number and its source where one exists.
- Earned mentions: reviews, buying guides and press in your market, which assistants cite more than your own pages.
- Measure across repeated prompts per assistant, as the share of model guide explains.
The Buyer Journey Check shows where your shop stands on each step, and Autobasket Discover runs the work and the measurement as a retainer.