What real GEO looks like — versus what you're being sold
The AI-visibility market sells two different products: work, and dashboards about work. Both call themselves GEO. One changes what AI says about your business; the other tells you, monthly, that nothing has changed. Here's how to tell them apart before you pay.
ASN Intelligence · July 2026 · every number below has a named source
Why this suddenly matters
In BrightLocal's 2026 consumer survey, 45% of consumers said they ask AI tools for local business recommendations — up from 6% a year earlier. Bain estimates roughly 60% of searches now end without a click to any website. The buying decision increasingly happens inside the answer, before your website ever loads. That's the real shift — and it's also why a gold rush of vendors appeared overnight.
The two products being sold as one
What you're buying
What it actually does
What changes
Tracking (most "AI visibility" software)
Asks AI engines about your brand on a schedule and charts the mentions.
Your awareness of the problem. Not the answer AI gives.
Auto-content ("GEO on autopilot")
Generates articles onto your own site; some trade links among their own customers.
Your page count. AI engines cite third-party sources for recommendations — your own blog mostly isn't the input.
Placement (the actual work)
Earns and fixes the things AI reads: reviews, directories, community presence, machine-readable pages, corroborated facts.
The answer itself. This is the part that requires humans and takes weeks.
None of these categories is a scam by definition. A tracking dashboard is honest if it's sold as a thermometer. The problem is thermometers sold as treatment — a monthly report proving you're still invisible, at a price that implies someone is fixing it.
What the evidence actually supports
Corroboration wins. In the peer-reviewed GEO study (Aggarwal et al., KDD 2024), adding quotations lifted AI visibility +41%, statistics +33%, and cited sources +28%. AI repeats what it can verify.
Freshness compounds. SE Ranking's 2025 analysis: content updated within 3 months averages 6 AI citations vs 3.9 for content over two years old.
Magic files don't. Ahrefs analyzed 137,000 domains and found 97% of llms.txt files received zero AI-bot traffic. If a vendor's headline deliverable is a special file, you now know what you're buying.
The visitors are worth more. Adobe's data shows AI-referred visitors convert 54% better than average — which is exactly why the answer slot is worth fighting for.
The five questions that expose a vendor in minutes
1. "Show me a baseline and a delta for a real client." Real work has a starting measurement and a change you can verify. A logo wall is not a delta.
2. "What exactly do you place, and where?" If the answer never leaves your own website, they can't move the sources AI actually cites.
3. "Will you take my competitor next month?" If yes, you're funding a machine that works against you the moment a rival subscribes.
4. "How many times do you sample each AI prompt?" AI answers change run to run. A screenshot is an anecdote; repeated sampling with intervals is a measurement.
5. "What happens when it doesn't work?" Honest shops publish misses. If every case study is a triumph, you're reading marketing, not measurement.
Our own answer to question 1: every engagement starts with a recorded, timestamped baseline scan, and every delta is reported against it — not against adjectives. We also test our own methods in registered, controlled experiments; the full results and measurement history publish when each trial concludes, in either direction. See the research page for the experiment registry.
What the real work looks like, week by week
Unglamorous, verifiable, and mostly off your own website: completing and correcting the profiles AI reads · earning and answering reviews · fixing pages so the text actually exists in the server-rendered HTML (most AI crawlers do not run JavaScript) · publishing answer-first pages with sourced facts · keeping name, address, phone, services, and claims identical everywhere · re-measuring the same prompts on a schedule and writing down what changed. That's it. That's the product. Anyone who can't describe their week in those terms is selling the dashboard.
Was this useful? (one tap, anonymous)
Who we are: ASN Intelligence is a done-for-you AI-visibility agency — the placement column of the table above. This guide is genuinely useful without hiring anyone; if you want your baseline measured, the free scan at asncheck.com is the same engine we use for clients. · More ASN research