Why does it matter that AI answers do not name you?
Buyers used to type a search and pick from ten links. Some buyers now ask an assistant — ChatGPT, Perplexity, Claude, Gemini — “which CRM suits a 12-person electrical firm?” and read one written answer. If your name is not in that answer, you are not on the list the buyer carries into the next conversation.
The clearest public number on how this changes behavior comes from Pew Research Center, which studied the browsing of U.S. adults in March 2025. When a Google results page showed an AI summary, people clicked a traditional result link in 8% of visits; when it did not, they clicked in 15% of visits. Read it as a measurement of Google in one month, not a law — but it shows why being named inside the answer now matters as much as being the link under it. The study is on Pew Research Center’s site.
What does nobody outside the AI companies actually know?
Most of what is written about “getting cited by AI” is guesswork presented as a formula. The companies behind the assistants do not publish how they choose which sites to name, and the answers change from one question to the next, and from one day to the next. Claims such as “question-form headings make you twice as likely to be cited” or “nine in ten brands get zero mentions” often come from a single consultancy’s own tests. We could not trace them to a source we could check, so we have left them out.
What is published is Google’s own advice for its AI features. Its guide to AI features and your website says there are no extra requirements and no special optimizations, and lists the ordinary things: let the page be crawled and indexed, link to it from your other pages, put the important content in text, keep the page pleasant to use, and keep any structured data in step with what visitors can see. That is a short list, and it is the right place to start.
Which fundamentals are worth doing?
These are low-risk because each one also helps a human reader. None of them is a guaranteed route into an AI answer.
| Practice | Why it helps | Where to start |
|---|---|---|
| Answer the question in the first lines | A reader (and a machine reading for a quotable answer) gets the point at once | Put a two-sentence answer above the first heading of every post |
| Let the page be found | A page that is blocked or not indexed cannot be named by anyone | Check robots rules, the sitemap and Search Console |
| Say what you are in one plain sentence | An assistant that cannot describe you in a line will describe you vaguely, or wrongly | Write the sentence, then use the same words on the home, pricing and about pages |
| Publish prices and limits in words | “Contact us” gives an assistant nothing to repeat; a stated limit stops it inventing a promise | List each plan, what it needs and what it does not do |
| Keep structured data honest | Google asks that markup match the visible page | Mark up only what the page really shows |
| Be described by other sites | Independent mentions are evidence the assistant can weigh | Give reviewers, partners and local associations an accurate fact sheet |
| Date your facts | Prices and features change; a dated page is easier to trust than an undated one | Write “as of October 2026” beside anything that can change |
What does a CRM’s own description need to say?
The most common reason a vendor is described badly is that its own pages never settle on one description. A usable one has five parts: what it is, who it is for, what it costs, what is extra, and what it does not do.
Here is how we write ours. Operations Link is operations software for the trades, trucking and the firms that serve them: clients, jobs, quotes, invoices, projects, team, email, chat, video meetings and a client portal in every plan. Essentials is $79 and Professional is $129 per person a month, billed in US dollars; Enterprise is a conversation. Fifteen days free, no credit card. Services such as Fleet, Field operations, Payroll, HR, Bookkeeping, Nova, Vera and the phone system are added separately on Professional and Enterprise. And the limits are said where they apply: there is no public API, pay runs calculate for Ontario, Canada today, phone numbers are Canadian today, and the books live in the company’s own QuickBooks Online. You can read the details on the pricing page and the platform page.
The limits are not an apology. They are the lines an assistant is most likely to get wrong if you leave them out, and a wrong line repeated in a thousand answers is worse than a missing one.
Why do honest comparison pages help?
Buyers ask assistants for comparisons: “Operations Link or Jobber?”, “HubSpot or something built for the trades?” If the only comparison text on the web is written by one side, the answer will be thin or slanted. A fair comparison names what each product does, links to the other company’s own page, says when the other product is the better choice, and carries a date. Ours are on the comparison pages, including HubSpot and Salesforce. They are written that way because it is the only kind worth quoting.
How do you find out what AI says about you?
Ask. Once a month, put the same ten questions to the assistants your buyers use — “what does X cost?”, “what are X’s limits?”, “best software for a small HVAC company?” — and write down what comes back, with the date. Answers vary between runs, so look for patterns, not single results. Where an answer is wrong, fix the page it most likely learned from, and say the correct thing in the plainest sentence you can. A paid tracking tool can do the asking for you, but it only reports; it does not change what the assistant says.
What is a sensible 30-day plan?
- Days 1–3. Write the one-sentence description and the list of limits. Use the same wording on the home, pricing and about pages.
- Days 4–7. Check that every public page is indexed, in the sitemap and linked from another page. Put a short answer at the top of each post.
- Days 8–14. Write the comparison pages you would want a buyer to find: facts only, each linked to the other company’s page, each dated.
- Days 15–21. Write the posts that answer the questions buyers actually ask, in the words they use. Reuse real questions from sales calls and support.
- Days 22–27. Send an accurate fact sheet to the people who might describe you: customers who like you, trade associations, reviewers.
- Days 28–30. Ask the ten questions again. Write down what changed and what is still wrong.
Nothing here promises a mention within a set number of days. It makes sure that when an assistant does look for you, it finds something true.
Frequently asked questions
Why isn’t my CRM cited when buyers ask ChatGPT for recommendations?
There is no published rule, but three causes are common and fixable. Your own pages may not describe the product in one clear sentence. Few other sites may describe it at all. And your pages may skip the plain facts an assistant can repeat, such as price, limits and who it is for. Start with the first and the third; they are in your hands.
How is “not being cited in ChatGPT” different from “not ranking on Google”?
A search results page shows several links and the reader chooses. An assistant writes one answer and names a few products inside it, and the reader may never click. You can rank well and still go unnamed, or be named without a click. Both are worth tracking separately.
What is the most useful thing a CRM can do this week?
Write down, in one paragraph, what the product is, who it is for, what it costs and what it does not do, and make the home, pricing and about pages say the same thing. It costs nothing and removes the most common source of wrong answers.
How do ChatGPT, Perplexity, Claude and Gemini decide who to cite?
Their makers do not publish the method, and it differs between products and over time. Treat any exact formula, such as a fixed multiplier for a heading style, with suspicion. The safest assumption is that clear, checkable pages and independent mentions help, and that nothing here is guaranteed.
What schema markup do AI assistants actually use?
Google’s guide says no special markup is needed for its AI features and that any structured data you use should match the visible text. Ordinary markup for articles, products and organizations is still reasonable for other reasons, as long as it describes only what the page really shows.
Does an AI-visibility tracker fix this?
No. A tracker reports whether and how assistants mention you. It cannot change the answer. It is useful once your pages are clear, because it shows whether the work is landing; before that it mostly confirms a gap you already know about.
How long does it take to start being cited?
Nobody can promise a timeline. Pages must be crawled, other sites must mention you, and assistants update on their own schedules. Expect weeks to months, and judge progress by whether the answers become more accurate, not only by whether your name appears.