Generative engine optimization research and sources · 44

Origin of the term GEO. Citing sources, adding quotations and statistics increased visibility in generative engine answers by up to 40%. Tested on a simulated engine (gpt-3.5-turbo) and Perplexity.ai.

OAI-SearchBot surfaces sites in ChatGPT search; sites that block it are not shown in ChatGPT search answers. GPTBot is for training. ChatGPT-User acts on user requests.

ChatGPT ranks search results using multiple factors; placement is not guaranteed. It sometimes partners with other search providers.

ChatGPT search uses third-party search providers and content from partners; any website can choose to appear in ChatGPT search.

Cloudflare, WAFs, CAPTCHAs and JS challenges can block crawlers with a 403. OpenAI publishes its search bot IP list.

ChatGPT Ads are available in 89 countries; ads appear only on the Free and Go plans; Plus, Pro and Enterprise stay ad-free. Ads are labeled, separate from answers and do not influence them.

Country-by-country self-service availability for ChatGPT Ads Manager; Turkey is listed as available. The legal entity that advertises and is billed must be based in a listed country.

ChatGPT has more than 1 billion weekly active users. ChatGPT Ads reached a $1B annualized run rate; self-service Ads Manager opened in India, Europe, the Middle East and North Africa; CPC and outcome-optimized bidding, Pixel and Conversions API.

No additional requirements or special optimizations for AI Overviews and AI Mode; a page must be indexed and eligible for a snippet. AI Mode may use query fan-out.

Optimizing for generative AI search is still SEO. Google Search ignores llms.txt files. No need to chunk content; structured data is not required. Use the generative AI report in Search Console.

A dedicated Search Console report for AI Overviews and AI Mode impressions. Rolled out to all sites on August 31, 2026.

The report shows impressions only (no clicks, CTR or position). Several results from one site in a single AI feature count as one impression.

A URL-prefix property supports many verification methods; a Domain property is verified via DNS only and covers all subdomains.

See a URL’s index status, test the live version and request indexing.

Sitemap formats and how to submit them to Google.

Shows where new and returning visitors come from: Reports → Acquisition → Traffic acquisition.

Google-Extended controls use for Gemini training and grounding; it does not affect inclusion in Google Search.

AI Performance in Bing Webmaster Tools: how often you are cited in Microsoft Copilot and Bing AI summaries, cited pages and grounding queries.

Grounding queries are the key phrases the AI used when retrieving content. The report does not measure ranking or authority.

Intents, Topics, Citation Share and Compare added. Citation Share is an observational metric, not a ranking.

Clear headings, Q&A pairs, lists, tables and schema help. Walls of text, content hidden in tabs, PDFs and text inside images hurt.

Keyword Research, up to 16 months of search performance data, Site Scan and importing sites from Google Search Console.

Bing recommends using both tools together.

One submission is shared with all IndexNow-enabled engines (Bing, Yandex, Naver, Seznam, Yep, Amazon). Google is not listed. No guarantee of immediate indexing.

Tracks visits from ChatGPT, Claude, Gemini, Copilot and Perplexity as a separate channel.

ClaudeBot is for training, Claude-User for user requests, Claude-SearchBot for search quality. Blocking Claude-SearchBot may reduce search visibility.

Brave Search and TurboPuffer are listed as web search subprocessors for Claude.

PerplexityBot crawls to surface sites in Perplexity results and is not used for model training. Perplexity-User acts on user requests.

AI traffic is split into Search, Agent and Training. Since September 15, 2026, new domains block Training and Agent bots only on pages with ads; Search stays allowed.

Shows which AI services access your site; allow/block rules per crawler. Available on all plans.

Managed robots.txt blocks training bots (GPTBot, ClaudeBot, Google-Extended, etc.); search bots are not on the list.

Cloudflare recommends starting new projects on Workers.

Static files are uploaded from the assets directory in your Wrangler config and served from Cloudflare’s network.

Requests to static assets are free and unlimited.

Free plan: 20,000 files per version, 25 MiB per file.

Every Worker gets a free workers.dev address.

A proposal for a Markdown summary of a site for agents. A proposal, not a standard.

Reports John Mueller (Google) saying in June 2025 that no AI system currently uses llms.txt. Second-hand report.

76% of pages cited in AI Overviews rank in Google’s top 10 (1.9M citations).

About 12% of URLs cited by AI assistants rank in Google’s top 10 for the same prompt (15,000 prompts). For ChatGPT it is 8%.

Branded web mentions show the strongest correlation with AI Overview brand visibility (0.664). Correlation, not causation.

AI traffic to US retail sites rose 235% year over year in Jan–May 2026; AI visitors converted 40% better than other channels.

AI referrals are under 1% of traffic but sign up at higher rates (1,277 domains).

Most major AI crawlers do not execute JavaScript (except Gemini and AppleBot). Important content should be server-rendered HTML.