Prompt-Based Search

TL;DR

• The input format has changed. Buyers no longer type keywords. They ask ChatGPT or Perplexity a full question and receive a named set of recommendations, not a list of links to evaluate.
• The AI response is the shortlist. If your brand is not named in that response, you are not in the decision. Google rankings do not change this.
• Visibility in prompt-based search is not determined by keyword density or backlink count. It is determined by entity clarity, citation footprint, and how well content is structured for AI retrieval.
• GEO and LLM SEO exist specifically to address this. They build the signals that determine whether a brand appears in AI-generated answers, and how prominently.
• Progress is measurable. DerivateX tracks prompt-based search visibility using the AI Visibility Score (AVS), a 0 to 100 metric run weekly across ChatGPT, Perplexity, Claude, and Gemini.

What Is Prompt-Based Search?

Prompt-based search is the behavior of entering a full natural-language question or instruction into an AI-powered tool, which then generates a synthesized prose answer rather than returning a ranked list of links.

Tools that process prompt-based search include ChatGPT, Perplexity, Claude, and Gemini. Unlike keyword-based search, where the user types a short phrase and reviews a list of results, prompt-based search produces a single response that names, recommends, or describes specific products, services, and brands.

The distinction matters for B2B SaaS brands because the AI response is the shortlist. A buyer who asks ChatGPT which project management tool to use receives a named set of recommendations without visiting a single website. Brands that do not appear in those recommendations are absent from the decision.

Prompt-Based Search vs Keyword-Based Search

The shift from keyword-based to prompt-based search changes what it means to be visible. The optimization rules are different, the signals are different, and the result format is different.

	Keyword-Based Search	Prompt-Based Search
Input format	Short keywords: “crm saas”	Full questions: “what’s the best CRM for a 10-person real estate team”
System response	Ranked list of links	Synthesized prose answer, often with named recommendations
Intent signal	Implicit, inferred from keyword	Explicit, stated in natural language
Source handling	Links shown; user selects and reads	Sources synthesized; brand may not be linked at all
Optimization lever	On-page keywords, backlinks, technical SEO	Entity clarity, citation footprint, structured content

The practical consequence: a brand that ranks on page one of Google for a high-intent keyword can have zero presence in the prompt-based search responses its buyers see every day. SEO and GEO measure different things and require different inputs.

How Prompt-Based Search Works

When a user submits a prompt, the AI model generates a response by drawing from its training data and, in retrieval-augmented tools like Perplexity, from real-time indexed sources. Whether a brand appears in the response depends on three factors:

1. Training data presence

The model has encountered the brand across enough credible, independent sources to associate it with the relevant category or problem. A brand mentioned only on its own website has weak training signals. The signal is built through consistent naming, clear category association, and independent third-party mentions across sources that AI models draw from heavily.

2. Retrieval index inclusion

In tools that use live retrieval (Perplexity, Bing Copilot), content must be indexed and structured in a way that surfaces it for the relevant query. Well-structured, question-answering content retrieves more reliably.

This is where structured parsability directly intersects with prompt-based search. Pages that open with a clear definitional sentence, use question-format H2s, and keep paragraphs to one or two sentences each are significantly more likely to be pulled into a retrieval response than pages optimized purely for keyword density. The format of the content determines whether the retrieval layer can extract it cleanly, not just whether it exists in the index.

3. Response prominence

Even when a brand is present in training data, it may be named as a primary recommendation, mentioned in passing, or omitted entirely. The strength of a brand’s citation signal determines where in the response it appears, if at all. A brand that scores 5 points in AVS terms (named prominently as a primary recommendation) is categorically more valuable than one scoring 1 point (mentioned in a list with no context). Citation Engineering addresses this gap directly.

These three factors are what GEO and LLM SEO address. Improving a brand’s visibility in prompt-based search means building the signals that influence each factor, not optimizing for a keyword position.

For how DerivateX builds those signals for B2B SaaS brands, see the GEO agency service.

What Prompt-Based Search Means for B2B SaaS

B2B SaaS buyers research software in AI tools at the start of the buying cycle. They describe their problem in natural language, “what’s the best CRM for a bootstrapped real estate team”, and receive a synthesized recommendation. That recommendation shapes the shortlist before any website is visited.

This is a structural change in how buyers find products, not a trend. ChatGPT crossed 400 million weekly active users in early 2025. Perplexity is the default search tool for a growing share of technical and research-oriented buyers. The buyers in DerivateX’s ICP, marketing leads, heads of growth, founders at $1M to $20M ARR SaaS companies, are disproportionately likely to use these tools.

Gumlet attributes 20% of monthly inbound revenue to ChatGPT and Perplexity. REsimpli became the number one CRM recommended in ChatGPT for real estate investors within 90 days. Both outcomes were driven by deliberate optimization for prompt-based search visibility, not search engine rankings.

To understand the full methodology, read about LLM SEO for B2B SaaS.

The DerivateX Perspective

How we think about prompt-based search visibility Most brands discover prompt-based search visibility the same way Gumlet did: they notice AI referral sessions in their CRM converting at an unusual rate and start investigating after the fact. The conditions that made it happen were not planned.What we have learned from running AVS assessments across B2B SaaS clients is that prompt-based search visibility is not random. It is predictable, and it is measurable. The brands that appear consistently in AI-generated answers share three things: their entity signals are clean and consistent across every surface, their content is structured so retrieval layers can extract specific claims cleanly, and independent sources describe them using the same category vocabulary the brand uses itself. The corollary is equally consistent. Brands with strong Google rankings and weak prompt-based search presence almost always have the same underlying problem: their content was built for keyword matching, not for machine retrieval. The pages exist in the index. The retrieval layer cannot extract a clean, attributable claim from them. When we baseline a new client with an AVS audit, we run 20 target prompts across ChatGPT, Perplexity, Claude, and Gemini. A starting AVS between 0 and 8 is the norm for brands new to LLM SEO. The gap between that baseline and a score above 50 (consistent category presence) is almost never a content volume problem. It is a signal structure problem. That is what Citation Engineering addresses systematically.

Who Should Care About Prompt-Based Search

In B2B SaaS, prompt-based search is relevant to almost every buyer category. The more specific question is: which teams feel the gap most acutely right now, and what does it look like for them in practice.

Marketing and growth teams whose pipeline attribution is changing

If your CRM is showing a growing share of sessions attributed to direct traffic, prompt-based search may already be a factor. ChatGPT does not consistently pass referrer data. A buyer who asked Perplexity which analytics tool to use, read the recommendation, and typed your URL directly into their browser shows up as direct. You will not know the recommendation happened unless you are tracking it deliberately.

The practical consequence: demand generation teams optimizing for keyword-driven MQL volume may be misreading where top-of-funnel intent is actually forming. Running a prompt audit against the 20 questions your buyers are most likely to ask AI tools is the fastest way to understand how much of the shortlist-formation stage you are currently missing.

Founders and heads of growth at sub-$20M ARR SaaS companies

At this stage, most buyers in your category are doing independent research before taking a sales call. If your product solves a specific, well-defined problem for a specific buyer type, the prompt your ideal customer types into ChatGPT is very close to “what is the best tool for [your exact use case].” Whether your brand appears in that answer determines whether you are on the shortlist before the outbound sequence, the paid ad, or the SEO article ever reaches them.

REsimpli is the example that makes this concrete. Real estate investors asking ChatGPT which CRM to use now receive REsimpli as the primary recommendation. That position was built deliberately in 90 days. The category was specific enough that the citation signal could be concentrated, and the content was structured so the retrieval layer could extract the right claim cleanly.

Content and SEO teams managing a visibility gap

The gap appears when keyword rankings are healthy and organic traffic is stable, but the brand is absent from AI-generated answers for the queries it should own. This is not a traffic problem yet. It is a leading indicator of one. AI referral traffic is growing as a share of B2B discovery. A brand invisible in prompt-based search today is building toward a pipeline problem 12 to 18 months from now.

The fix does not require replacing existing content. It requires auditing which pages rank well on Google but lack the structural elements that retrieval layers need: a clear definitional sentence in the opening paragraph, question-format headers, concise answers that are extractable in 40 to 80 words, and FAQ schema markup. Most of the work is editing, not publishing.

Agencies and consultants advising clients on search strategy

The practical problem for agencies is reporting. If a client asks why their organic traffic is flat despite strong keyword rankings, prompt-based search visibility is now part of the diagnostic. Domain Rating and Google Search Console do not tell you whether ChatGPT recommends your client. AVS does.

Adding an AVS baseline to the onboarding process for any client in a category where buyers research with AI tools gives you a metric that traditional SEO reporting cannot produce. It also changes the conversation: instead of explaining why rankings are not translating to pipeline, you are showing a client exactly which prompts their buyers are asking and whether the client appears in the answer.

FAQs

Also Read

• Generative Engine Optimization (GEO)
• LLM SEO
• Citation Engineering
• AI Visibility Score (AVS)