AI citations are not business recommendations

An AI citation means an answer links to a source. A business recommendation means the answer offers a provider as a suitable choice. You can receive either without the other, so a report that counts citations should not quietly call them recommendations.

Consider a fictional gate-repair business. An assistant links to its guide while explaining how to spot a worn hinge. The guide may help answer the question, but the answer has not necessarily suggested hiring that business. It may not even establish that the business serves the reader's area.

Reverse the example: an assistant offers the company as a local option and cites an independent directory. That is a recommendation supported through an outside source, even though the company's own website received no citation.

Read the sentence before counting the success

A name appearing in an answer is only the beginning of the review.

"Example Company repairs gates" describes a service. "Example Company may suit this repair in your area" presents an option. "Example Company does not handle this kind of repair" mentions the company but is not a positive recommendation. These are fictional statements, not findings about a real provider.

Check that the answer identifies the right business. Similar names, old locations, and former owners can produce a plausible-looking match that should not count as a success.

Then inspect the link. Does the cited page support the statement next to it? Does it describe the right service and location? If you cannot open the source or verify its support, mark that part unresolved.

OpenAI's Search guidance explicitly warns that results and citations can be incomplete, outdated, or incorrect. A visible link is a reason to check evidence, not a substitute for checking it.

Keep a few separate fields

For a practical review, record whether the answer:

• Mentions the correctly identified business.
• Offers it as a suitable provider.
• Cites the business's own site in support of a relevant statement.
• Cites an outside source that supports a statement about the business.

These are LeadSpark's suggested observation definitions, not platform-issued ranking metrics. One answer can qualify for several fields. Do not add the owned-site and outside-source totals to get "answers with any supporting citation" without removing the overlap.

Also note whether you supplied the business name in the question. A correct answer to "What does my company do?" checks recognition. It does not show the assistant would have found the company when asked for an unnamed provider.

What if our guide gets cited, but our business is not recommended?

First check the question. An educational question may not ask for providers at all.

If the question does call for a provider, compare the answer's requirements with your public information. Are your actual services, geography, and limits clear? Does your site provide evidence for relevant qualifications? Do not assume more articles will solve missing service information.

Those checks identify possible improvements. They do not prove why a particular answer omitted you.

Does no citation mean the assistant never found us?

No. The final answer does not expose the whole retrieval process. Record "no visible citation" rather than "never retrieved." If search use itself is not clear, leave it unknown.

Is a recommendation a lead?

Not by itself. A recommendation in your test is an observed answer. A visit, inquiry, and sale are later outcomes that need their own records. Even a real customer may have encountered the business through several channels.

The search-readiness guide covers what you can improve. The measurement guide explains how to repeat observations without inflating the result.

If a visibility report leaves you unsure what to fix, bring the full answers and their sources to a System Audit conversation. Custom work is scoped after that review, not sold as a guaranteed position in someone else's answer.