SEO and GEO in 2026: how to rank on Google and get named in AI answers

How search and AI answers work

AI answers now sit on top of Google Search and inside the assistants people use. All of them are built from pages that search engines can read.

1. Step 1 · Crawl and indexCan they read you?Googlebot, OAI-SearchBot, PerplexityBot and Claude-SearchBot fetch your pages. robots.txt decides for each bot separately.
2. Step 2 · RetrieveWhich pages fit?Google runs several related searches across subtopics, called query fan-out, and collects supporting pages.
3. Step 3 · AnswerWho gets named?The answer cites or links a handful of sources. Pages that are not used get nothing.

Google ran two core updates in 2026 by the end of September, starting 27 March and 21 May, each taking 12 days, plus spam updates in March, June, August and September. Check the Search Status Dashboard dates before you react to a drop.

Strategy 1: technical health in an afternoon

If a search engine cannot crawl, index and quote a page, no ranking and no AI answer can use it.

1. Indexing. Search Console › Pages shows what is indexed and why other pages are not. Fix your key pages first.
2. Snippets. A page marked noindex, nosnippet or with a tight max-snippet cannot be quoted in AI Overviews. Check it is on purpose.
3. Crawlers. robots.txt decides bot by bot. Blocking a training bot is a separate choice from blocking a search bot.
4. Internal links. Every important page is linked from related pages with descriptive words, a few clicks from the home page.
5. Page experience. Core Web Vitals on real visits: LCP within 2.5 s, INP within 200 ms, CLS below 0.1.
6. Structured data. It must match the visible text. It helps Google understand a page; no study shows it raises AI citations.

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

User-agent: ClaudeBot
Disallow: /

OpenAI says robots.txt changes take about 24 hours to apply. User-triggered fetchers such as ChatGPT-User and Perplexity-User may not follow robots.txt.

Strategy 2: content that ranks and gets used

Google rewards pages made to help people. AI answers draw on pages that answer clearly and cover the whole question.

Google's own test is who, how and why. Who made it: an author with real experience in the topic. How it was made: how you tested or sourced it, and where AI helped if readers would expect to know. Why it exists: to help people, not to rank.

1. Plan the fan-out. List the questions around a topic: basics, costs, comparisons, problems. One page per real question.
2. Answer first. Open with two sentences that answer the question. Then give the detail, the steps and the proof.
3. Update for real. Add new facts and examples. AI assistants tend to cite fresher pages. Never change only the date.
4. Use AI with care. AI drafting is allowed. Mass-producing pages that add nothing can count as scaled content abuse, a spam policy.

The same page opening, before and after, for a fictional brand. E-E-A-T is not a score, there is no 2,500-word rule, and Google does not penalise AI-written text that helps people.

Strategy 3: proof of expertise

Sourced numbers, named voices and first-hand evidence make a page easier to trust, and in one lab study, easier for AI answers to use.

The researchers rewrote pages and measured how much of a generative engine's answer drew on them. The caveats: 2023-era models, and it measures a share of the answer text, not clicks. Results varied by topic.

• A number with its source. Your own data first: tests, orders, survey results. Otherwise a primary source, linked.
• A named voice. Quote an expert or a customer by name and role, with permission.
• First-hand evidence. Photos you took, results you measured, a method you can show. What only you can publish.

That Ahrefs correlation (75,000 brands, May 2025) suggests mentions matter more than links, but a correlation is not proof. Earn mentions the honest way: be quoted in your category's roundups and comparisons, publish data others cite, and collect reviews on sites people trust.

Strategy 4: local search

Google ranks local results on relevance, distance and prominence. You control two of the three.

1. Verify the profile. An unverified Business Profile cannot be fully managed.
2. Fill every field. Category, services, hours (holiday hours too), attributes and, for shops, products.
3. Add photos and videos. Real photos of the place, the team and the work, refreshed as things change.
4. Ask for reviews, reply to all. Ask every happy customer. Reply to every review, good or bad, specifically.

"Post every week or you drop" and "thirty days without activity hurts your rank" appear nowhere on Google's ranking page. What moves prominence: more genuine reviews, a good rating, replies that show you listen, and mentions from local sites and press.

Strategy 5: GEO, get named in AI answers

Generative engine optimisation is mostly good SEO plus proof. Spend effort only where there is evidence.

The anatomy of a page AI answers can use

1. The question as the heading, in the words people use, not a slogan.
2. A two-sentence answer, before any story or context.
3. A proof block: a sourced number, a named quote, a primary link.
4. The detail, by subtopic: one section per follow-up question from the fan-out.
5. Who, how and why: the author, the method, and the date of the last real update.

Four myths to answer when someone sends them: "rank first in Bing and ChatGPT will cite you" (no OpenAI page explains how ChatGPT picks sources), "block Google-Extended to leave AI Overviews" (it does not), "post on Reddit and AI will cite you" (studies show Reddit is cited often, not that posting gets you cited), and "this tracker gives your exact AI rank" (trackers sample prompts, and answers change by person, place and day).

Strategy 6: measure rankings and AI mentions

The monthly AI panel

1. Fix 20 prompts: the questions customers ask before buying, general, comparison, local and your brand by name.
2. Run them the same way in ChatGPT, Gemini, Perplexity and Claude, from the same place and account type each month.
3. Score and compare: mentioned, cited with a link, competitors named. Track the rates month by month, not single answers.

In a randomised study of 1,065 US desktop users, organic clicks fell 39.8% when an AI Overview appeared. Judge pages on impressions, AI mentions and conversions too, not clicks alone.

Run it with Claude

Claude reads your search numbers through connectors, and none of these can edit your website: Windsor.ai (Search Console and GA4, in Claude's directory), Semrush and Ahrefs (their own official connectors, which need paid plans), and Google's experimental Analytics MCP server for Claude Code. Google has not published a Search Console MCP server; read the code of any community server before you connect one.

Rewrite <URL> so people and AI systems can use it. Read it first. Keep
our facts; improve how they are shown.
1. Open with a two-sentence answer to the page's main question.
2. Add proof where we have it: a statistic with its source, a quote
   from a named expert or customer, a link to the primary source.
   Where I must supply the real number, write <NEED DATA>.
3. Add who, how and why: who wrote it and why they know, how we tested
   or sourced it, and the date of the last real update.
4. Keep only claims we can prove.
Show the new version, then a list of what changed and why.

The PDF has the rest: a site health audit, a topic plan for the whole fan-out, titles from real queries, an internal links plan, a local profile audit, the monthly AI visibility panel, a robots.txt check, a monthly report and a content refresh list.

Your first 90 days

• Month 1 · fix what blocks you. Run the audit and the crawler check, fix indexing on key pages, complete the Business Profile and connect Claude.
• Month 2 · improve what ranks. Rewrite your five most valuable pages answer-first with proof, and write new titles for low-CTR pages.
• Month 3 · grow what is missing. Plan the fan-out, write the missing pages, start the AI panel and add internal links.

Written by

Anisha Dogra

Anisha runs performance marketing and AI systems for brands at RAAY Creative, the AI Lab and marketing studio she co-founded in New Delhi. She writes these guides from the work the studio does every week, and checks every fact against its source before it goes up.

Talk to us about this