What is AEO visibility, actually?

What is the AEO visibility formula?

AEO visibility = (number of buyer queries where an AI engine mentioned your brand) ÷ (number of buyer queries run) × 100

That's it. A percentage. Sliced by engine (ChatGPT, Perplexity, Gemini, Brave), by buyer-journey phase (awareness, consideration, evaluation, decision), and by location if you sell into geographies.

Lower is bad. Higher is good. The rate of change matters more than the absolute number.

Why is visibility percentage the right AEO metric?

There are five candidate AEO metrics floating around the industry. Most of them are wrong, in interesting ways:

1. "Times we appear in AI search results" - counts mentions, but a single mention in a 50-prompt sample isn't the same as a single mention in a 5-prompt sample. The denominator matters.
2. "AI ranking position" - there is no ranking position. AI engines return one answer, not ten links. "Position" is a SEO-era concept that doesn't translate.
3. "Citation count" - counts URLs from your domain that AI engines reference. Useful as a sub-metric, but it doesn't measure whether your brand is being recommended; it measures whether your content is being used as source material. Different things.
4. "Sentiment score" - also useful as a sub-metric. But a single negative mention is worse than zero mentions, and a single positive mention with no others is barely a signal. Sentiment without volume is noise.
5. Visibility percentage - the share of queries where you're in the answer at all. The clearest leading indicator of whether AI engines view you as part of the conversation in your category.

Visibility is the metric that tracks the actual question every founder cares about: "when buyers ask AI assistants about our category, are we even part of the answer?"

What four signals determine AEO visibility?

You move AEO visibility by moving four underlying signals. Not three, not seven. These four:

1. Citation pattern (which sources AI engines pull from)

When ChatGPT answers a query about CRM tools, it pulls from a small number of authoritative sources: G2, Capterra, vendor blogs, comparison sites. If those sources mention you, you appear. If they don't, you don't.

The work: get cited by the sources AI engines already trust. This is closer to traditional PR and earned-media work than to keyword optimisation.

2. Page-level extractability (can AI engines pull from your own pages)

Your own pages need to be written for AI extraction. Specifically: the answer to a likely buyer question needs to be in the first 50–100 words of the page (cross-encoder rerankers truncate at ~512 tokens), the page needs to use Hearst-pattern sentence structures ("X is a type of Y", "X such as Y, Z, and W"), claims need to be sourced (AI engines weight content that itself cites), and entity coherence needs to be intact (sameAs, JSON-LD, brand-name consistency across the site).

The work: rewrite the lead-answer paragraph. Add JSON-LD. Get the first 512 tokens to do meaningful work. This is content engineering, not link building.

3. Entity coherence (AI engines have to know who you are)

AI engines disambiguate brands the same way Wikipedia does - through sameAs links to authoritative IDs (Wikidata, LinkedIn, Crunchbase, Common Crawl), through consistent brand naming across the visible web, and through Hearst-pattern co-occurrence with the category they recognise you as part of.

If AI engines confuse your brand with a same-named competitor, your visibility will look chaotic - high one week, zero the next, depending on which entity the reranker resolved to. The work: claim authoritative profiles, link them via sameAs, keep brand name consistent.

4. Buyer-query coverage (the prompts you're tracking matter)

Visibility against the wrong prompts is a vanity metric. If you sell project-management software for engineering teams, tracking visibility on "best project management software" tells you about a category you can't credibly win. Tracking it on "project management for embedded firmware teams" tells you about the segment you can.

The work: make the prompt list narrow enough to be winnable, and broad enough to cover the buyer journey from awareness to decision.

How AEO is different from SEO

We have a longer piece on this (AEO vs SEO), but the short version: SEO measures ranking on the keyword × page axis. AEO measures recognition on the query × passage × engine axis. The signals overlap (good page health helps both) but the optimisation axes are different.

A page that ranks #1 in Google for "best CRM" can be invisible to ChatGPT for the same question because:

• The answer is buried in section 4 (cross-encoder rerankers don't see that far)
• The brand name doesn't appear in the lead paragraph
• The competitor's blog cites G2 (and G2 mentions them by name in their listing) while your blog cites no one

AEO and SEO solve overlapping problems with different toolkits.

What is the difference between AEO, GEO, AIO, and LEO?

The term GEO - Generative Engine Optimization - was coined in a 2023 academic paper by researchers at Princeton, UMass Amherst, and Georgia Tech. Their study found that structured content changes (adding statistics, sourced claims, and direct quotations) increased brand mention rates in AI-generated responses by up to 40% compared to unstructured prose.

Three other acronyms have been proposed for the same problem space:

• AIO - AI Optimisation. Identical to AEO; the term hasn't won.
• GEO - Generative Engine Optimisation. Functionally identical, popularised by a 2023 Northwestern paper; same metric, same techniques.
• LEO - LLM Engine Optimisation. Same concept, less common.

We use AEO because it's the clearest about what's being optimised: visibility on Answer Engines (engines that produce a single answer, not a list of links). The acronym fight is mostly noise; the underlying problem is the same.

Why this metric should be a board-deck slide

Three reasons AEO visibility deserves the same attention as your traffic chart:

1. Buyer journeys are reshaping around AI assistants. The "I'll Google it" reflex is being replaced by "I'll ask ChatGPT" for a growing share of high-intent research queries. SEO-only visibility is a partial picture.

2. AEO visibility moves faster than SEO visibility. Google rankings move on the order of weeks to months. AEO visibility moves day to day as AI engines refresh their retrieval indexes and competitors publish content. You can ship a fix on Monday and see the visibility number move by Friday.

3. AEO visibility is more predictive of revenue than SEO traffic. A query asked of an AI assistant is high-intent - the buyer has already framed their problem, they're asking for a recommendation. SEO traffic includes a lot of accidental clicks, comparison shoppers, students writing essays. AI-engine queries skew toward decision-stage intent.

Founders should be tracking it. Marketers should be reporting it. Boards should be asking about it.

What good AEO visibility looks like

Rules of thumb from cohort data we see across the platform:

• Below 15% - your brand is not yet in the AI engines' answer set for your category. You're invisible.
• 15–35% - you're starting to show up. Likely concentrated on a few prompts (probably branded queries - buyers asking about you specifically).
• 35–60% - you're a recognised player in your category. AI engines reach for you when listing options. Your job from here is consistency across engines and depth across the buyer journey.
• 60–80% - you're a category leader for the prompts you've selected. Worth widening the prompt scope to harder queries; you've maxed your existing set.
• 80%+ - you're dominant. Worth asking whether your prompt list is too narrow (i.e., too easy to dominate).

These aren't industry-wide; they're observations from across the brands we track. We're publishing a public benchmark by category once cohorts pass the min-N gate.

What to do about it

Three things in order:

1. Measure baseline. Run a scan. See your current visibility per engine, per buyer-journey phase. The first scan reveals which engines you're invisible on and which prompts you're losing to which competitors. Free tier covers this.

2. Pick the highest-ROI prompt to fix. Look at prompts where:

• A competitor was mentioned and you weren't (someone is winning the conversation)
• The query is decision-stage (highest revenue intent)
• The answer was on Perplexity or ChatGPT (highest-traffic engines)

3. Ship the content. Each AEO recommendation in CiteAgentic ships with a 1500-word AI fix prompt - drop into Claude Code, get a publishable blog post or page rewrite. Ship it. Re-scan in seven days. Watch the visibility number move.

That's the loop. Repeat weekly.