A business can be indexed by Google, listed in every directory, and named in a ChatGPT answer, and still lose the job. The listing proves the business exists. It gives a buyer no reason to pick it.

That gap is the subject of this issue. The video below walks through five steps. This article explains the logic underneath them, where the evidence is solid, and where it runs out.

The short version: AI search rewards the business that answers the buyer's choosing question on its own page, backs the answer with evidence, watches how AI describes it, and can trace the result to a quote request.

Owners tend to talk about "showing up in AI" as a single result. It is at least five separate outcomes, and each one depends on the one before it.

Indexed. A search system has stored your page. Google's documentation on AI features says a page is eligible to appear as a supporting link in AI Overviews or AI Mode only when it is indexed and eligible to show with a snippet, with no extra technical requirements. The same page states that meeting every requirement does not guarantee indexing or serving.

Mentioned. An AI answer includes your name. A mention can be neutral, wrong, or buried in a list of eight.

Cited. The answer links to your page as a source. A citation means the system used your content to build the answer. It does not mean the answer favors you.

Recommended. The answer points the buyer toward you for their situation. This is the outcome that moves a decision.

Qualified inquiry. A real buyer contacts you about work you do, in a place you serve. This is the outcome that pays.

A business can be strong at the first two and absent from the last three. The process in the video targets two gaps: the one between cited and recommended, and the one between recommended and an inquiry you can trace.

Take a hypothetical company. Call it Harbor Clean, a commercial cleaning firm serving medical and dental offices in one metro area. Every detail about Harbor Clean in this article is invented for illustration.

Harbor Clean's front desk and estimator hear the same questions every week. Some are requests for information: "How often does a dental operatory get disinfected?" Others are choosing questions: "Which cleaning company handles medical offices and follows infection-control protocols?" or "Who can clean after hours without disrupting patient schedules?"

The difference matters. An information question can be answered well by a health department page, a trade association, or a supplier blog. AI systems have many good sources for it, and the buyer has not decided to hire anyone yet. A choosing question is a provider decision. The buyer is comparing who fits their constraints.

Collecting these questions is not a brainstorm. The best sources are records the business already keeps: inbound emails, call notes, quote request forms, sales call recordings, and the objections that come up before a contract is signed. Harbor Clean would sort them into the two groups and pick three choosing questions tied to its most valuable work.

Many service pages describe the service. Few answer the question that decides the hire.

For the after-hours question, Harbor Clean's medical-office page would state the answer in its first lines: the hours crews work, how building access and alarm codes are handled, how the office confirms the work was done, and what happens if a crew misses a night. Each claim would carry evidence a buyer or a machine can check: the training or certification its crews hold, the products it uses and why, the service area by town, and links to reviews or references from medical offices specifically.

That evidence layer separates a page that gets cited from a page that supports a recommendation. An AI system can quote a definition from anyone. To suggest a provider for a specific situation, it relies on specific, consistent, verifiable statements that match the buyer's constraints.

The platforms' own guidance stays close to fundamentals. Google lists making important content available as text, keeping structured data consistent with the visible page, and keeping Business Profile details current. It also states that no special AI text files or schema are required to appear in its AI features. OpenAI's publisher FAQ says any public website can appear in ChatGPT search, as long as its OAI-SearchBot crawler is not blocked in robots.txt.

What remains unknown: none of these companies publishes how a recommendation is chosen among eligible sources. A clear, evidenced answer improves the odds. It does not guarantee inclusion, and a promise of guaranteed placement is a promise the platforms themselves do not make.

Once the page is live, Harbor Clean would ask its three choosing questions in the AI tools its buyers use, on a set schedule, and log the results. Was the company named? Was it cited? Was it recommended? How was it described, and which competitors appeared beside it?

Two first-party reports help. In June 2026, Google added generative AI performance reports to Search Console, showing how often a site's URLs appear in AI Overviews, AI Mode, and AI features in Discover. Those reports count impressions. They do not show what the reader did next. In February 2026, Bing Webmaster Tools introduced AI Performance, which shows how often a site is cited across Microsoft Copilot and Bing's AI summaries, and which URLs are referenced. It reports citation activity, not clicks.

Manual checks have limits too. AI answers shift with phrasing, location, account history, and time. Five checks in one week are a sample, not a ranking. The useful signal is the pattern across weeks: whether Harbor Clean moves from absent, to mentioned, to described accurately, to recommended for the after-hours question.

The last step is the one skipped most often. A page that earns a recommendation still turns into revenue only if the buyer can request a quote easily, and only if the business can see that it happened.

Harbor Clean's medical-office page would end with a short quote form tied to that exact question, with fields for facility type, square footage, and preferred cleaning window. The form would record the page it came from. The intake call would ask one plain question: how did you find us?

Some AI traffic identifies itself. OpenAI states that ChatGPT adds utm_source=chatgpt.com to referral links, so those visits show up as their own source in analytics. Much of the effect will not identify itself. Pew Research Center analyzed 68,879 Google searches by 900 U.S. adults in March 2025. When an AI summary appeared, users clicked a traditional result in 8% of visits, compared with 15% when no summary appeared. They clicked a link inside the summary itself in 1% of visits.

That study has limits worth naming. It covers Google only, predates the wider rollout of AI Mode, and does not measure what buyers do later. A buyer can read an AI answer, close the tab, and search the business by name two days later. That inquiry looks like branded search or direct traffic, with no trace of the AI answer that started it.

So the three numbers stay separate. An AI mention, a website visit, and a signed contract are different events. Harbor Clean would track each one, connect them where the data allows, and treat a self-reported "an AI tool suggested you" as useful but incomplete evidence.

LinkedIn is a practical place to test choosing questions with real buyers. A post that answers "who can clean a dental office after hours" shows quickly whether office managers recognize the problem and how they phrase it. Nothing in the platform documentation reviewed for this issue shows that posting on LinkedIn causes AI systems to recommend a business. Use it to reach people and sharpen the answer. Then publish the final answer on the website, where the business controls it.

Upcoming issues go deeper on what this one only touches: measuring return, choosing platforms, tool categories, lead scoring, and connecting the whole path to a CRM.