ChatGPT began recommending this B2B scheduling SaaS startup after the company rewrote three basic facts - who it is, whom it competes with, and what it costs - as passages an AI engine could extract without filling in missing context. A Google ranking jump was not the cause. After ninety days, inquiries from AI conversations across ChatGPT, Perplexity, and Google AI overview grew by 58%, while organic search traffic remained almost flat. This article explains what we changed, why we used that sequence, and which decisions another team can reuse.
Starting Point: Page-Two Google Rankings and No AI Visibility
The client was a 20-person Taiwan startup selling scheduling and attendance SaaS for restaurant and retail chains. Its sales team kept hearing the same thing on first calls: "I asked ChatGPT, and it recommended A and B." The client never appeared on the list. Its website was solid, the blog was active, and core Google keywords ranked on pages 1 to 2. Yet AI engines never cited the company for queries such as "scheduling software in Taiwan" or "attendance systems for restaurant chains." Search could see the company; AI could not. This is the most common gap we have encountered over the past year.
Step One: Measure How AI Describes the Brand Before Writing
The usual instinct is to publish more articles. We did not. In the first week, AI visibility monitoring ran 20 real buying questions through ChatGPT, Perplexity, Gemini, and Google AI Overviews. We recorded whether the company appeared, whether each description was accurate, and which source pages the engines cited. That replaced the vague feeling of low exposure with a baseline the team could track question by question.
The baseline was worse than expected, which made the next actions clearer. Four gaps accounted for most of the problem:
- Only three of the twenty questions mentioned the company, and each contained its brand name. Situational comparison and recommendation queries produced no mentions.
- AI misclassified the product as a basic time-clock tool and missed its strongest feature, automated scheduling, because the website never stated that distinction in an extractable definition.
- A competitor had a clear "Compare with X" page that Perplexity cited repeatedly. The startup had no comparison pages, leaving competitors to define the category and the terms of comparison.
- Pricing said only "Contact us," leaving AI with nothing to cite. The company was therefore omitted from answers about prices and plans.
Implementation: Add Three Types of Extractable Content
We did not publish at scale. We rewrote or added three types of pages, each designed around a structure AI could cite cleanly. The first was a definition passage at the top of the product page: "This scheduling and attendance SaaS is built for restaurant and retail chains and automatically creates shifts from revenue forecasts and labor-law working-hour rules." One sentence established the category, audience, and differentiator without asking the model to infer anything.
The second type was a comparison page. We selected the three products that prospects mentioned most often and wrote an honest page for each, including the areas where the other product was stronger. That candor made Perplexity more willing to treat the page as a neutral source rather than an advertisement. The third type turned pricing, company size, and common questions into short Q&A passages that could each stand alone.
We made two technical changes in parallel. Product, pricing, and FAQ content received the appropriate Schema markup, and we confirmed that the key pages did not depend too heavily on client-side rendering. AI crawlers could then retrieve complete text instead of an empty shell. Both details are easy to overlook, but they determine whether an engine can access the content at all.
Results after 90 days
After launch, we reran the same set of 20 questions every two weeks. The trend was clear. By the sixth week, the company began appearing in comparison and recommendation answers. By the tenth week, when asked which scheduling systems suited Taiwan restaurant chains, ChatGPT named the startup among three recommendations and correctly identified automated scheduling as its differentiator.
- Across 20 situational buying questions, the share that mentioned the company correctly rose from 15% to 70%.
- Website inquiries attributed to AI conversations rose 58% in 90 days, including form submissions and callers who said they had found the company through ChatGPT.
- The sales team reported that AI-referred prospects understood the positioning more accurately. First-call friction fell because the team no longer had to explain that the product was more than a time clock.
- During the same period, Google organic search traffic stayed within 5% of its starting level, indicating that the growth came from AI visibility rather than traditional SEO.
The biggest change was not the traffic number. Customers already understood what we did when they called. We used to spend 10 minutes clarifying our positioning; now, prospects arrive through an AI recommendation and begin with the questions that matter.— Marketing Lead at the Startup
Three Lessons to Reuse from This Case
First, measure before you write. Test real buying questions to see how AI currently describes the brand, then address the actual gap in its definition, comparison, or pricing instead of publishing on instinct. Second, make comparison pages honest. Acknowledging a competitor's strengths can earn AI's trust and citations; hiding them keeps the brand out of more comparison answers. Third, measure visibility separately from search traffic. This case barely registered in Google Analytics organic search, so the old metrics alone would have made the gains invisible.
The method works across industries, but every company has a different gap and sequence of fixes. To see how AI engines currently describe your brand and which missing signals cause them to skip it, book a 30-minute GEO diagnostic session. We will test the buying questions from your own market and show you the gaps.
