What is AEO / GEO? A plain guide for B2B companies

AEO, GEO, AI SEO: same job, different names

GEO comes from a 2023 research paper that tested how content changes affect visibility in generative search answers [1]. AEO is older: it first described optimizing for featured snippets and voice assistants, and is now used for AI assistants too. You will also see “AI SEO” and “AI search visibility”. They describe the same goal, so we use AEO / GEO together.

How an AI assistant decides what to say

When a buyer asks “Which companies make benchtop XRF analysers for alloy sorting?”, three things shape the answer:

1. What the model already knows. Well-known brands are in the model's memory before any search happens.
2. A live web search. For commercial questions, assistants usually search the web, then write the answer from the pages they retrieve and cite.
3. Which index they search. Google's AI Overviews and AI Mode use Google's index. Microsoft Copilot uses Bing. ChatGPT search and Perplexity run their own crawlers alongside other search sources. Ranking well on Google alone does not cover all of them.

This is why smaller or newer suppliers struggle. A 2026 study of startups found that ChatGPT recognised 99.4% of them when asked by name, but surfaced only 3.32% when users asked open discovery questions [2]. Being known by name and being recommended are different problems.

What moves the answer for B2B companies

1. Let AI search crawlers in

Each AI company publishes the user agents of its crawlers [3][4][5]. Search and retrieval crawlers, such as OAI-SearchBot, ChatGPT-User, PerplexityBot, Claude-SearchBot and Bingbot, need to reach your pages for you to be cited. Training crawlers such as GPTBot are a separate choice, and OpenAI documents the two settings as independent [3]. Check your robots.txt and your CDN's bot settings, which can block AI crawlers without anyone noticing.

Put key content in the HTML the server sends. Many AI crawlers do not run JavaScript [6], so specs that only appear after a script runs, or only inside a PDF, may never be read.

2. State your facts plainly on your own site

Assistants repeat whatever is easiest to find. If your detection limit, operating range, compatibility, certifications, lead times or regions served are missing or buried, they will be filled in from a distributor listing, an old review, or a guess. One clear, current page per product line, written from your datasheet, is the most controllable fix there is.

3. Be present on the pages assistants cite

For category questions, assistants lean on third-party pages: directories, distributors, trade media, reviews, forums and video. In a study of 75,000 brands, branded web mentions correlated with AI visibility at 0.66 to 0.71, and YouTube mentions at about 0.74 [7]. That is correlation, not proof of cause, and bigger brands get mentioned more anyway. Still, the direction is clear: being written about by others tends to matter more than adding pages to your own site.

4. Keep doing the search basics

Assistants that search the web start from search results, so indexing and rankings still matter. Make sure you are verified and indexed in both Google Search Console and Bing Webmaster Tools, and submit a sitemap to both. IndexNow can speed up discovery on Bing and other engines [8].

5. Answer the questions buyers actually ask

Your sales inbox, site search and support tickets contain the exact questions buyers bring to AI assistants. A page that answers one of them directly, with real numbers, is more useful to a buyer and to an assistant than another general product overview.

What doesn't work, or isn't worth paying for

• llms.txt files. Google says its Search, including its AI features, does not use them [9], and an analysis of 137,000 sites found that 97% of llms.txt files never got read [10]. Harmless, but not a strategy.
• Markup as a shortcut. Structured data helps search engines understand pages, but there is no special markup that gets you into AI answers [9][11].
• Mass-produced AI pages. Publishing many low-value pages to manipulate rankings is “scaled content abuse” under Google's spam policies [12].
• Self-serving “best suppliers” lists, fake reviews and hidden prompts. They break search policies and consumer protection rules, and they damage the trust you're trying to build.

How to measure it honestly

AI answers are not stable. SparkToro found less than a 1 in 100 chance that ChatGPT or Google's AI would return the same list of brands in any two responses to the same prompt [13]. So a screenshot of one answer proves very little, in either direction.

Measure with a fixed set of questions, run several times on each assistant, and report how often you appear. Keep assistants separate. Add your own data: Bing Webmaster Tools now reports AI citations [14], and your analytics can show referral visits from AI assistants. Our method page describes how we do it.

A 30-minute self-check

1. Open yourdomain.com/robots.txt. Is OAI-SearchBot, ChatGPT-User, PerplexityBot, Claude-SearchBot or Bingbot disallowed?
2. If you use Cloudflare or another CDN, check its bot and AI crawler settings. They can block crawlers that robots.txt allows.
3. Open a key product page and use “View page source”. Are the specs in the HTML, or only in a PDF or loaded by script?
4. Check that your site is verified in Bing Webmaster Tools and that your sitemap has been submitted.
5. Ask ChatGPT, Perplexity and Google AI Mode five questions a buyer would ask, without your company name, three times each. Write down who gets named and which pages are cited.
6. Ask each one “What does [your company] make?” and compare the answer with your own datasheet.
7. Add “How did you hear about us?” to your inquiry form, with AI assistants as an option.

If steps 1 to 4 turn up problems, fix those first. They are the cheapest wins. If you'd like us to run the checks for you, properly sampled, ask for a free AI visibility check.