TL;DR / Key Takeaways
- AEO, GEO and AI visibility are one category with three vocabularies — search demand splits ~4,400/2,400/1,600 monthly across the names.
- Measuring one buyer question across six engines costs ~24.5¢ at API prices; scraper-based collection is ~16× cheaper — vendor pricing is dashboard margin, not data cost.
- ChatGPT's API browsed the web on only ~74% of our production calls; grounded and ungrounded mention counts are different products.
- Mention counts from answer corpora are floors, never rankings — we measured a corpus-count-of-one site getting cited in 3 of 4 live answers.
The category has three names because it grew from three directions at once. "Generative engine optimization" comes from the 2023 academic paper that coined the term (Aggarwal et al., KDD 2024), which showed content changes can lift visibility in AI answers by up to 40%. "Answer engine optimization" came up through the SEO industry. "AI visibility" is what the tool vendors converged on. Search demand is split across all three — our own July 2026 pull of Google Ads volumes puts "generative engine optimization" at ~4,400 searches a month, "answer engine optimization" at ~2,400, "ai visibility tools" at ~1,600 and "llm seo" at ~880. Same buyers, four vocabularies. Whatever you call it, the tools do two things: measure where AI systems mention you, and tell you what to change.
The market, in tiers
We keep a verified price-and-feature comparison of the five biggest monitors that we update as vendors move; the short version of the landscape:
- Enterprise: Profound is the reference brand, selling to large marketing orgs. Scrunch was acquired by Sitecore on June 3, 2026 — Bloomberg put the price around $225 million — which is the category's first real exit and the clearest signal that AI visibility is becoming a line item in enterprise CMS budgets.
- Mid-market: Peec AI (raised a $21M Series A in 2026), Otterly.AI, and the AI-visibility toolkits inside Semrush and SE Ranking. These run scheduled prompts across engines and chart your share of answers against competitors.
- Entry and free: single-purpose checkers and trackers, including our own free AI visibility checker — paste a URL, see in ten seconds whether AI crawlers can read your site, whether your pages serve text without JavaScript, your authority profile, and how many AI answers in a commercial corpus already cite your domain. No signup.
For a menu across monitors, agencies and DIY — with what each actually costs — see our cost breakdown and the full tools roundup.
What it costs to measure — from our own invoices
We run a six-engine visibility monitor in production, so unlike the listicles, we can tell you what the measuring itself costs. Across 925 monitored runs in July 2026, the vendor-billed cost of asking one buyer question across all six engines was 24.5¢:
| Engine | Cost per grounded answer | The catch nobody prints |
|---|---|---|
| Claude (Anthropic API) | 10.1¢ | The most expensive engine to measure |
| ChatGPT (OpenAI API) | 6.7¢ | Browses the web on only ~74% of calls — the other 26% answer from memory at ~0.3¢ |
| Grok (xAI API) | 3.5¢ | Reports billed cost in the response itself — the only vendor that does |
| Perplexity API | 3.0¢ | The per-request search fee dominates; tokens are almost free |
| Google AI Overviews (via SERP API) | 1.0¢ | Returned AI Overview data on only 49% of our calls |
| Gemini (API) | 0.2¢ | Effectively free to measure |
Two consequences. First, a $99-a-month monitor tracking 20 prompts weekly across five engines carries maybe $5–10 of engine cost — you are paying for the dashboard, the diffing and the alerting, which is fine, but negotiate like you know that. Second, scraper-based measurement (reading the ChatGPT consumer UI rather than the API) costs as little as 0.4¢ per answer with citations — 16× cheaper than the API — which is how the cheap tier of this market exists at all.
Enjoying this? Get one like it in your inbox each morning.
one email a day · unsubscribe in two clicks · no third-party tracking
The ~74% problem: grounded vs ungrounded mentions
The single most important technical question about any ChatGPT visibility number: was web search on? The API decides per-call whether to browse. In our production data it browsed about 74% of the time even when instructed to; the rest of the answers come from the model's training memory — a different distribution of brands, frozen months ago. A vendor averaging the two into one "ChatGPT mentions" number is averaging two different products. The consumer ChatGPT UI browses far more consistently, which is why serious monitors are moving to scraper-based collection of the real UI rather than API calls that approximate it.
Counts are floors — the inversion we measured
Several tools sell mention counts from pre-collected corpora of AI answers. Useful data — we buy it too — but know what it is. When we compared corpus counts against live, web-grounded answers for the same buyer questions, the ranking inverted at both ends: one directory with a corpus count of one was cited in three of four live answers we ran, while another with a corpus count in the hundreds was cited in none. A corpus is a sample of questions people already asked, not a census of what engines say today. Treat every mention count as a floor — "cited at least this often" — never as a ranking, and never let a vendor sell you a zero as proof you're invisible.
Five questions to ask any AI visibility vendor
- Grounded or not? Are your ChatGPT numbers from web-search-on answers, and can I filter to only those?
- Which engines, exactly? ChatGPT, Perplexity, Gemini, Claude, Grok and Google AI Overviews answer differently; a two-engine average is not "AI visibility."
- Prompt-bound or demand-bound? Do you only track prompts I write (what I thought to ask), a corpus of real user questions (what people actually asked), or both?
- Can I see the answers? The row-level answer with its cited sources is the product; a score without the receipts is a vibe.
- What moves the number? If the tool only measures, who does the fixing — and what did it cost the last customer who moved from unmentioned to mentioned?
Where to start
Start free and mechanical: run your domain through the AI visibility checker to catch the plumbing failures (blocked crawlers, JavaScript-only pages, missing structured data) that no paid monitor will fix for you. Then, if you buy a monitor, buy it with the five questions above. And a disclosure, since we've named half the market: Stork also sells visibility work — Stork Wire publishes editorial coverage on real sites in your market, on the pages engines read when they decide who to recommend — so weight our verdicts here the way you'd weight any vendor's, receipts included.