AEO and GEO for Beginners: The Plain-English Starting Guide

AEO (answer engine optimization) and GEO (generative engine optimization) are two names for the same job: getting your company named, and ideally linked, when a buyer asks ChatGPT, Claude, Perplexity, Gemini or Google's AI Overviews a question you should be the answer to. SEO still sits underneath it. What changes is the unit of work, the scoreboard, and who gets to vouch for you.

This is the plain-English starting point. It covers what is different from SEO, what is not, how an AI engine decides which brands to name, and the five things to do first, in order, with templates you can copy today.

What AEO and GEO mean, without the jargon

You will hear three acronyms. Treat them as one discipline with three historical labels.

Agencies argue about the boundaries. For a B2B company it barely matters, because the tasks overlap almost completely. If you want the longer comparison, read SEO vs AEO vs GEO and AEO vs SEO. This guide uses "AEO" for the whole practice.

What actually changes versus SEO

Six things change. Each one changes how you plan the work.

1. The unit is a prompt, not a keyword

A buyer types "CLM software" into Google. The same buyer asks ChatGPT: "We are a 300-person manufacturer on SAP Ariba. Which contract management tools integrate natively and how are they priced?" The prompt carries industry, size, stack and constraints. You now optimize for situations, not two-word phrases. The article on prompt intent versus keyword intent goes deeper.

2. There are far fewer slots

A results page shows ten organic links. An AI answer cites a handful of sources. In Orbit Media's study of 13,184 citations across 1,765 answers, Claude averaged 3.6 sources per answer and ChatGPT 4.5. Gemini averaged 8.1 and Perplexity 19.2. If you are not in that small set, the buyer may never see you.

3. Being mentioned and being cited are separate wins

A mention means your brand name appears in the answer text. A citation means your URL appears as a source. The same Orbit Media data shows the gap: Claude mentioned the tracked B2B brands in 70% of answers but cited them in 40%. For ChatGPT the figures were 56% and 47%. You need to track both.

4. Other people's pages carry a lot of the weight

AI engines lean on third parties to confirm who is credible. Orbit Media found Clutch was Claude's single most-cited third-party domain, while Perplexity's most-cited domains were LinkedIn, YouTube and Reddit. All four models cited the same domain for a query in just 1.7% of cases, so each engine has its own favorite sources. Your website alone is not enough.

5. Google rankings do not transfer one to one

Ranking well helps, but it is not the same scoreboard. Orbit Media measured 10% to 30% overlap between AI-cited domains and Google's top 10, and only 1% to 5% at exact URL level for Gemini and ChatGPT. The rankings and AI visibility correlation article covers the nuance.

6. Answers change from run to run

Ask the same question twice and you can get different brands. Week to week, Orbit Media saw Claude keep only 30% of its cited URLs, while Perplexity was the most stable at 65%. Never judge your position from a single screenshot. See why LLM answers keep changing.

What stays the same

Google is direct about this. Its guidance on AI features says there are no extra requirements or special optimizations for AI Overviews and AI Mode. A page must be indexed and eligible to show in Search with a snippet. Google also says you do not need new machine-readable AI files or special schema to appear.

So the basics still decide a lot:

• Crawlable, indexable pages with the important text in the HTML
• Clear, specific writing from people who know the subject
• Accurate structured data that matches what the page says
• Internal links that make your key pages easy to find
• A reputation that exists outside your own domain

AEO adds prompt research, answer-first page design, entity consistency and off-site corroboration on top of that foundation. It does not replace it.

How an AI engine decides who to name

The mechanics differ by engine, but the sequence is similar enough to plan around:

1. Interpret the prompt. The engine works out what is being asked and often splits it into several sub-searches. This is called query fan-out, covered in query fan-out optimization.
2. Retrieve candidate pages. It pulls pages from a search index, which is why crawler access matters.
3. Read passages, not whole sites. It favors passages that answer a sub-question cleanly and specifically.
4. Write the answer and attach sources. Brands that several sources agree on are safer to name.

The engines lean on different source types. This table summarizes Orbit Media's findings. The study tracked 72 prompts for three B2B brands over about three months, so treat it as directional.

The practical takeaway for beginners: publish clear primary pages for ChatGPT-style retrieval, and build a third-party footprint (reviews, listicles, LinkedIn, video) for the rest. The mechanics are explained in how answer engines work.

The five things to do first

Do these in order. Each step feeds the next one.

1. Write down 30 real buyer prompts

Do not start from a keyword tool. Start from what buyers actually say. The best sources are sales call recordings, RFP questions, win and loss notes, the objections your SDRs hear, and reviews of competitors on G2 or Clutch.

Spread the 30 prompts across five families so you see the whole buying journey:

PROMPT SET v1 (30 prompts)

Category (8)      best [category] for [industry/size/region]
                  e.g. best ServiceNow partner for mid-market healthcare
Problem (6)       how do I [fix problem] without [constraint]
                  e.g. how do I cut contract cycle time without new headcount
Comparison (6)    [competitor A] vs [competitor B] for [use case]
Alternatives (5)  alternatives to [market leader] for [segment]
Validation (5)    is [your brand] good for [use case]
                  what does [your brand] do / cost / integrate with

For each prompt record: family, persona, funnel stage, priority (1-3)

The full method for sourcing and weighting prompts is in how to build an AI visibility prompt set.

2. Find out where you stand today

Run each prompt in ChatGPT, Claude, Perplexity and Gemini. That is 120 answers. For each answer, record four things in a spreadsheet:

• Is your brand named? (yes or no)
• Is your website linked as a source? (yes or no)
• Is what it says about you accurate? (yes, partly, no)
• Which competitors and which cited domains appear?

Then calculate two numbers. Mention rate is answers naming you divided by total answers. Citation rate is answers linking your site divided by total answers. Those two numbers are your baseline. When you are ready to do this properly, with repeat runs, competitor share of voice and a monthly cadence, follow how to benchmark your AI visibility.

3. Make sure AI crawlers can reach your site

Many B2B sites block AI crawlers by accident through a robots.txt rule, a CDN bot setting or a firewall. If the engines cannot fetch your pages, nothing else in this guide works. These are the user agents that matter, as described in each company's own documentation:

A safe starting point for a B2B marketing site allows the search-facing bots explicitly:

# robots.txt (search-facing AI bots allowed)
User-agent: OAI-SearchBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: *
Disallow: /admin/
Sitemap: https://www.example.com/sitemap.xml

Whether to allow training crawlers such as GPTBot and ClaudeBot is a business decision. The decision guide on blocking AI crawlers lays out the tradeoffs. Then check your CDN and firewall, because a clean robots.txt means little if the WAF serves bots a challenge page. See WAF blocking AI crawlers.

4. Rewrite five pages answer-first

Pick the five pages that should answer your highest-priority prompts. Usually that is the main service or product page, one industry page, one comparison page, a how-it-works or pricing-model page, and your strongest case study.

On each page, the first two or three sentences must answer the page's main question directly, with specifics: who it is for, what it does, what it integrates with, how it is priced, and proof. Here is the difference:

BEFORE
We are a passionate team of data experts helping organizations
transform their data journey with innovative solutions.

AFTER
Acme Data is a Snowflake Elite services partner that migrates
mid-market healthcare and insurance companies off Teradata and
on-prem SQL Server. Typical migrations run 12 to 20 weeks, and we
have completed 40 of them since 2021. We price fixed-fee per
migration phase.

The "after" version (a fictional company) gives an AI engine a passage it can lift and attribute. Use question-style H2s, tables for comparisons, and a visible last-updated date. The structural rules are in content structure for LLMs.

5. Get named on three trusted third-party pages

Go back to your baseline spreadsheet and list the domains the engines cited for your prompts. Those are the gatekeepers in your category. Pick three you can realistically appear on in the next 60 days:

• A complete Clutch profile with recent client reviews, if you sell services
• A G2 profile with recent reviews, if you sell software
• A category listicle that already gets cited, where you can request inclusion or a correction
• Your partner directory or marketplace listing, if you are in a vendor ecosystem

The full playbook is in listicles, Clutch and G2: the off-site assets that get B2B brands cited.

A worked example

Illustrative example. A 60-person Salesforce consulting partner in Texas runs the five steps.

• Prompts: 30 prompts, weighted toward "Salesforce partner for manufacturing" and "Revenue Cloud implementation partner" category prompts.
• Baseline: 120 answers. The firm is named in 6 (a 5% mention rate) and linked in 2 (under 2%). Three national SIs appear in most answers.
• Cited domains: a Clutch category page, two agency-written listicles and competitors' case study pages show up again and again.
• Access check: the CDN's bot protection returns a challenge page to AI user agents. The team fixes the rule in an hour.
• Pages: the manufacturing page, Revenue Cloud page and two case studies are rewritten answer-first, with named outcomes and a comparison table.
• Off-site: the firm asks ten clients for Clutch reviews and emails both listicle authors with a factual inclusion request.

Thirty days later they re-run the same 30 prompts with the same protocol and compare. They do not judge the result on one run, and they tie the prompts back to the deals sales is working.

Beginner mistakes to skip

• Treating llms.txt as the fix. llms.txt is a proposal by Jeremy Howard, and Google says you do not need AI text files to appear in its AI features. It is a low-priority extra. See does llms.txt work.
• Judging from one screenshot. Answers vary. Measure across runs and engines.
• Writing for robots. Fifty thin FAQ entries stuffed with phrases will not be cited. Specific, expert passages will.
• Ignoring brand prompts. "What does [your brand] cost" is often answered wrong. Accuracy is part of the job. See fixing wrong information in ChatGPT.
• Skipping question and comparison queries in Google. Seer Interactive found comparison queries triggered AI Overviews 95.4% of the time and question-format queries 85.9%. These are exactly the pages B2B teams tend to skip.
• Measuring AEO apart from pipeline. Mentions matter only if they reach buyers who book meetings. Add an AI assistant option to your "how did you hear about us" field from day one.

Ten terms you will hear

• Prompt set: the fixed list of buyer questions you track over time.
• Mention rate: the share of answers that name your brand.
• Citation rate: the share of answers that link to your domain as a source.
• Share of voice: your mentions divided by all mentions of the brands you track.
• Entity: your company as a distinct "thing" machines recognize, with consistent facts across the web. See entity SEO.
• Schema: structured data (JSON-LD) that states facts about a page or organization in machine-readable form.
• Query fan-out: an engine splitting one prompt into several sub-searches.
• Answer block: a short, self-contained passage that directly answers one question.
• Corroboration: third-party pages that confirm what you say about yourself.
• AI referral traffic: visits that arrive from AI engines, trackable in GA4 with a custom channel group. See the GA4 setup guide.

Where to go next

This guide is the first in a sequence. Work through them in order:

1. The AEO checklist for a B2B website: every technical, entity, schema and content check, with a pass or fail test.
2. How to benchmark your AI visibility: the scoring method and table template.
3. AEO in 90 days: a week-by-week plan starting from zero.
4. The off-site assets that get B2B brands cited.

If you sell into a specific vertical, the industry guides go further: system integrators, data consulting firms, cybersecurity and higher education, where our work with St. Martinus University grew visibility by roughly 3,600%.

For the full field guide, download the AEO and GEO playbook. If you would rather have someone run the baseline for you, our AEO, GEO and SEO team can start with a free audit.