ChatGPT has started to take the initiative to give the B2B platoon, SaaS, a novel to potential clients, not because its key word ranking was going to Google first, but because it changed the three things to "who we are, who we are, how much we are," to "the paragraph where the AI engine can clean up at a time without the need for brain fixes." Ninety days later, the number of queries from the AI dialogue (ChatGPT, Perplexity, Google AI overview) grew by 58%, while its natural search traffic remained almost static during the same period. This article breaks down what we actually did, why the order is so, and which judgments can be moved directly.
Start: Google ranks on the second page and AI disappears.
This is a 20-person Taiwan novel, which is done by locking food, food, retailing classes and attendance at SaaS. The trouble they had when they came to the house was very specific: business returns, and more and more potential clients say on the first phone, "I asked ChatGPT, which recommended A and B, and who never appeared on the list." Their official network is not bad, and their blogs are continuing, and Google's core words are on pages 1 to 2. The question is, the AI engine does not quote them at all when it answers questions such as "What kind of scheduling software is in Taiwan" "A attendance system suitable for food chaining". Searches can see, AI can't see, this is the most common gap we've had in almost a year.
The first step is not to write. It's to measure "Ai."
Most of the team's instinct is to produce a few more articles, we didn't. In the first week, Brand Radar put 20 real buying situations into the ChatGPT, Perplexity, Gemini and Google AI overviews, documenting three things: whether the company was mentioned, right or wrong, and which source pages AI cited. The value of this step is that it replaces the vague sense of "I don't think we're being exposed enough" with a baseline line that can be tracked by type.
Measuring the results is worse than expected, but it is better to do it. The gap is concentrated in four places:
- Of the twenty sets of questions, only three would mention the company and would appear when brand names were directly searched, and the situational comparisons and recommendatory questions would be completely zero.
- AI, speaking of their product location, mistook them as a purely time-ticking card tool, missing the strongest auto-teaming function -- because the official network never wrote this as a extractable definition.
- The competition has a clear page of the "Comparison with X" page, which has been quoted repeatedly by Perplexity; this new page has not been created, which is tantamount to giving the right to compare language to the opponent.
- The price was written as "Call us," and AI was not quoted, so it was skipped in any answer to the price, the program.
Do it: add three "AI extract units"
We don't have a big volume article, but we've rewritten and added three types of pages, each of which responds to the structure of the AI engine that can be most cleanly quoted. The first is the definitional paragraph: at the top of the product page, it is clear in one sentence: `This is a set of automatic shifts and attendances designed for chaining and retailing. SaaS, which can be automatically scheduled according to the forecast of operating rates and the rules of working hours of the Law,' to give full coverage to categories, perceptions, differences, and not allow AI to guess.
The second category is a comparative page. We picked three of the competitions they most often competed for at the sales field, each with an honest picture — where the other is actually stronger. The counterintuitive approach is precisely that of avoiding bad writing, which allows Perplexity to refer to the page as a neutral source of information, rather than as an ad. The third category is to recast the price formula, the applicable scale, and the commonly seen questions into a short paragraph in a question-and-answer format, each of which is individually available.
The technology adds two things: to put products, prices, FAQs on the corresponding Schema, and to make sure that these key pages are not over-described at the front end, allowing the reptiles to get complete text instead of shell. These two were often ignored and decided directly that AI couldn't catch you.
In 90 days.
We retest it every two weeks with the same set of 20 questions. In the sixth week, the problem of comparing and recommending began to arise. Their names; on the tenth week, ChatGPT, when asked about the Taiwan Food and Food Chaining System, listed them as one of the three recommended, correctly describing the difference between automatic shifts.
- Of the 20 purchase situational problems, the proportion correctly mentioned by AI rose from 15% to 70%.
- The official online query from the AI channel grew 58% in 90 days.
- The industry reported that this batch of queries from AI clearly had a better understanding of product location, that the cost of communication on the first call had dropped, and that it would not take time to explain that "we are not just a cardmaker."
- During the same period, Google's natural search traffic changed within 5% positive or negative, proving that the wave was growing from AI visibility, not traditional SEO.
The most touching is not traffic numbers, but clients already know what we do when they call. It used to take 10 minutes to clarify the location, and now it's recommended by AI, and it's important to start with.— The new marketing manager.
Three judgments that can be taken directly from this case.
First, first, first, first. Before the pen, the real buying question is how AI describes you now, and you know whether it's a definition, a price or not, rather than a feeling. Secondly, the page should be honest. Write about the strong points of the opponent in exchange for AI's trust and reference; hiding will only keep you absent on more similar issues. Third, visibility and search traffic need to be separated. The growth of this case is hardly reflected in Google Analytics' natural search figures, and if you look only at old indicators, you wonder why it never happened.
This approach is not tied to specific industries, but the gaps are different from one household to another. I want to know how the AI engine describes you now, and what's missing from the loop that you're skipping, and we're going to have a 30-minute GEO diagnosis, and we're going to use your own shopping situation to see where the gap is.
