AI assistants name three businesses. Be one of them.
Ask ChatGPT, Claude or Perplexity where to eat, stay or go and you get a shortlist of three or four names, not ten links. We work on the two things that decide who is on it: pages those crawlers can read, and independent sources that describe you accurately.
<h1>The Corner Room</h1>
<p>A twenty-four cover neighbourhood
restaurant in Fitzroy, Melbourne,
serving a daily-changing Modern Australian menu.
Open Wednesday to Sunday, dinner
only. Bar seats walk-in, tables by
reservation.</p>
Search has moved from lists to synthesis.
Traditional SEO optimized for a position in a list of links. Generative Engine Optimization works on the two things an assistant needs before it can name you: a page it can fetch and read without running JavaScript, and independent sources that describe you the same way you do.
- Optimises for
- A position in a list of ten links.
- Being one of the three or four names a synthesised answer contains.
- Counts
- Keyword coverage, backlink volume, rank tracking.
- Whether a crawler can fetch and read the page without JavaScript, and what sources you do not own say about you.
- Measured by
- Position, per keyword, daily.
- Share of answer across a frozen prompt set, sampled repeatedly because the same prompt does not return the same answer twice.
- Where it stops
- Zero-click results, where the answer is already on the page.
- A business nothing independent has been written about. We can fix the page; we cannot invent the coverage.
Four categories, and the queries that decide each.
We work only in hospitality, because the value is knowing which constraints decide a dining or accommodation query and which sources get cited for them. That knowledge does not transfer.
Specialty Cafés
Roaster, brew methods, wifi, seating, laptop policy. The most attribute-driven queries in hospitality.
High-End Restaurants
Cuisine, price per head, party size, dietary provision and booking policy decide who gets named.
Boutique Hotels
Room-level attribute depth and trip-intent language, against OTAs with far more content.
Tour Companies
Duration, difficulty, group size, language and weather policy, published per experience.
Why an assistant names three businesses and not yours.
A worked illustration of what an assistant returns for a category query, and the four things that have to be true before your name can appear in it. The states below are an example, not a measurement of any business.
What has to be true first (worked example)
-
Pass
Page readable without JavaScript Hours, address and the menu are in the HTML the crawler receives.
-
Pass
Answering crawlers allowed OAI-SearchBot, Claude-SearchBot, PerplexityBot and Bingbot all reach the page.
-
Partial
Independent mentions off-site Named in two local guides. The competitors in this answer appear in six.
-
Blocked
Answer near the top of the page The paragraph under the H1 is atmosphere. The stored snippet carries no facts.
An assistant answering this returns three roasters by name, each with a suburb, a house roast and opening hours attached: [Roaster A, South Melbourne], [Roaster B, CBD], [Roaster C, Prahran]. Venues whose pages it cannot fetch and read are not named at all, even when they are closer, cheaper or better reviewed.
This panel is a worked example, not a live model call and not an audit. The four checks are the prerequisites we work on, shown here in a plausible state for an independent venue. No business named or implied has been measured by NovaTechRay.
For the same four checks run against eight real businesses, with the method and the quoted evidence, see the Melbourne coffee teardown.
Four stages, and we will tell you if stage one is enough.
Four stages, set out in full on the methodology page. We tell you at the end of the first whether the rest is worth doing.
Baseline
We run a frozen prompt set across ChatGPT, Claude, Perplexity and Google AI Overviews, repeatedly rather than once, and fetch your site the way the crawlers do, with JavaScript disabled.
Retrievability
We move page content to the server so it survives with JavaScript off, unblock OAI-SearchBot, Claude-SearchBot, PerplexityBot and Bingbot in robots.txt and at the CDN, and clear the dead URLs those crawlers keep hitting.
Off-site mentions
We sample which sources the assistants actually cite for your category, get your details consistent across them, and work on being included in the guides that came back.
Measurement
We re-run a frozen prompt set on a fixed cadence, logged out, and report what moved along with how confident we are that it moved at all.
More of these on the FAQ, and the working behind them on the evidence page.
Understanding how Generative Engine Optimization differs from legacy digital marketing. Every claim on this site is sourced on the evidence page.
Generative Engine Optimization (GEO) is the work of making a business retrievable and corroborated. Retrievable means an AI crawler can fetch your page and read it without running JavaScript, and nothing in robots.txt or at your CDN is turning it away. Corroborated means sites you do not own describe you accurately. Those are the inputs an assistant has when it decides who to name.
Traditional SEO focuses on earning high placements on search engine results pages for specific keywords. GEO focuses on retrieval: whether a crawler can read the page at all, whether the answering bots are allowed through, and what independent sources say about you. A good deal of the underlying technical work is shared between the two.
No. There is no submission form and nothing you can file that puts you in an answer, so nobody can honestly promise placement. Google's Search Console now reports impressions in its AI features, but it shows no clicks, no position and no queries. What we can do is fix the mechanical prerequisites, work on the off-site mentions, and sample a fixed prompt set over time so you can see what moved.
Nobody can tell you honestly, because there is no index status to check and no platform publishes its retrieval cycle. What we can do is confirm the mechanical prerequisites are met, then sample a fixed set of prompts on a schedule so you see movement rather than take our word for it. Treat any specific timeline promise, ours included, as a sales tactic.
Ready to make your business discoverable to AI answers?
Book a 30 minute call. We will look at how ChatGPT, Claude and Perplexity currently answer questions in your category, and tell you whether there is anything worth doing.
Book a call