Veloce

What AI Search Engines Use to Name a Business

What AI Search Engines Use to Name a Business

AI search engines return a short list of businesses in response to a question, and most owners assume the businesses named have better websites. That is rarely what decided it. The engines do not read your site to build that answer. They read other people's pages about you, and which pages those are determines everything.

The model is not browsing your website

Large language models are trained on a snapshot of the web, not a live crawl of your homepage. Even when an engine has web search turned on, it does not start at your domain and read down the page. It sends a query to a search index, pulls a cluster of pages that look authoritative on the topic, reads those, and synthesises an answer. Your website is competing to be in that cluster. If it is not, the engine does not know your site exists for this particular question, regardless of how good your on page copy is.

This is why two businesses with nearly identical Google rankings can get very different results from AI answers. One has been written about across independent sources; the other has not.

What does count: third party pages that mention you specifically

The pages that actually feed AI answers tend to share a few properties. They are not your own. They are specific, not generic. And they are indexed quickly enough to be fresh when a user asks with web search enabled.

The pattern is credibility through repetition across independent sources. A mention only counts when the name, the category and the location appear together in a way that a model can connect to a real entity.

Why does it matter whether web search is on or off?

It matters because the two modes surface different problems. With web search off, the model answers from training data alone, which means you are dealing with a reputation problem that is months or years old. Listings that were thin or absent when the model was trained will not be fixed by a new blog post today. With web search on, the model fetches live pages, which means freshness matters a great deal and a recently published directory listing or editorial mention can move the needle within days.

Asking both ways is the only way to know which problem you have. If you appear with web search on but not without it, your reputation in the training data is weak and you need to build it over the coming months. If you do not appear either way, the source pages do not exist yet and that is the more urgent fix.

How does the engine decide which businesses to name?

It names the businesses that appear most consistently across the pages it read, weighted toward pages the model treats as editorially independent. Three factors tend to dominate.

First, volume of mentions across distinct domains. One very long article about you counts for less than several separate mentions on separate sites. Second, specificity of the mention. A page that names your business in the context of the exact question the user asked carries more weight than a generic directory listing that puts you in a broad category. Third, consistency of entity signals. If the same business name, category and suburb appear together across multiple sources, the model is more confident it is describing a real, operating business and not a defunct listing or a naming collision with a competitor.

This is also why similar names cause problems. If a competitor's domain contains your trading name as a word, some AI engines will attribute their mentions to you or yours to them. Clean entity signals across every source reduce this kind of bleed.

Which AI engines should you be measuring?

Veloce's AI Search module runs across four engines: ChatGPT, Gemini, Microsoft Copilot and Google's AI answers. Each matters for a different reason. Gemini sits inside Google's own ecosystem and draws on Google's index for its sourcing. Microsoft Copilot is built into Windows and Edge, which means a segment of users, particularly in business to business markets, reach it before they reach a standalone search engine. Google's AI Overviews appear directly in search results and their sourcing logic is closer to traditional SEO than the others, but they still pull from third party pages rather than your site alone.

Measuring one and ignoring the others leaves gaps. The source pages that feed one engine and the source pages that feed another overlap but are not identical, and the gap tells you where to build next. In a typical week, Veloce records 87 source pages across those four engines for a single site, which is enough data to see clearly which directories and publications are deciding your market.

What getting on the right source pages looks like in practice

The practical work is less glamorous than it sounds. It is finding which directories your market's AI answers are pulling from, getting your listing onto those directories with correct and specific information, earning editorial coverage on the sites the engines treat as authoritative for your category, and making sure your business entity is described consistently across all of them.

The trade off most owners face is time. Finding the source pages requires running real queries across multiple engines and recording every citation, not guessing which directories matter. Getting onto editorial sources takes either a relationship with a journalist or the kind of content that earns a mention from one. Neither happens in a week, but both compound over months.

One practical rule of thumb: if a directory page appears as a source citation in more than one AI engine's answer for your category, it is worth being on. If a directory page has never appeared as a source in any answer you can find, it probably will not help you regardless of how much traffic it claims to send.

A question worth asking any provider: how do you record the source pages?

Most tools in this space report a score, a percentage or a position. None of those tell you which pages the engine read to produce the answer. The only output that is actually useful is the list of source pages, ranked by how often they appear, because that list is the real map of where you need to build presence.

Veloce's AI Search module captures every page the engines cite when asked the questions your customers ask, then ranks those pages so you can see which sources are deciding your market. It asks each engine both with and without live web search, and records whether your business was named using a strict match that excludes competitors whose domains happen to contain your words. That distinction matters more than it sounds: a loose match inflates your apparent presence and sends the remediation effort in the wrong direction. The module costs $99 a month on its own, or is included with all four modules for $199 a month. There is a three day free trial with no card required, and any fixes made during those three days are yours to keep whether you continue or not.

Frequently asked questions

Do I need a good website to be named by AI search engines?

A working, indexed website helps establish that your business is a real entity, but it is rarely what gets you named. The engines build their answers from third party pages about you, not from your own domain. A business with a modest website and strong directory and editorial coverage will typically appear more often than a business with a polished site and no presence on the pages the engines read.

How long does it take to start appearing in AI answers?

It depends on which mode matters to you. With web search enabled, a fresh directory listing or editorial mention can be picked up within days of being indexed. In training data mode, the timeline is much longer because you are building the kind of presence that gets absorbed into a future model snapshot. Both are worth pursuing, but the fixes are different.

Which AI engine matters most for my business?

There is no single answer that applies to every market. Veloce checks four engines because the source pages each one reads are not identical, and the engine that matters most for a tradesperson in regional Australia may be different from the one that matters most for a B2B service provider in the United States. Measuring all four and comparing the source pages is the only way to know.

What is the difference between measuring AI visibility and measuring SEO?

SEO measures where your site ranks in a list of links. AI visibility measures whether your business name appears in a synthesised answer, and which third party pages the engine read to get there. The two overlap but are not the same, and improving one does not automatically improve the other.

Can I do this myself without paying for a service?

You can run queries manually across the four main engines and note down which businesses appear and which pages are cited. The difficulty is doing it consistently, across enough question variants, in both web search on and web search off modes, and recording 87 or more source pages per week in a format you can act on. Most owners find the time cost outweighs the subscription cost quickly.

All articles

Veloce