Quick Takeaways
- 1Prompt volume is estimated, not real. Large language models don't publish query data — what tools report as "prompt volume" is a modelled guess.
- 2AI responses are non‑deterministic. People phrase questions differently and LLMs use probabilistic decoding, meaning answers vary wildly.
- 3Rankings in AI tools are random. A SparkToro study found fewer than a 1‑in‑100 chance of identical brand lists across AI runs.
- 4Data sources are biased. Panels and API calls often reflect tech‑savvy users or API‑only prompts, not real human behaviour.
- 5Citation drift is massive. AI citation patterns change by dozens of percentage points month‑to‑month.
- 6Start with your ideal customer. Your audience's pain points and language are more reliable than vendor dashboards.
- 7Cluster and monitor intelligently. Organise prompts by intent, use tools directionally and schedule monthly monitoring.
Artificial intelligence is reshaping how people discover brands. Chatbots like ChatGPT, Perplexity and Google's AI Overviews answer questions directly instead of sending users to search result pages. This shift has given birth to Generative‑Engine Optimisation (GEO) — the art of influencing how AI answers recommend your business. If you've already heard the hype: just find the highest‑volume AI prompts and crank out content that matches them. For many marketers, that sounds familiar — after all, search engine optimisation was built on analysing keyword volume.
But there's a problem. LLMs don't work like search engines. There is no public AI search volume. Tools that promise prompt data are modelling and estimating, not reporting actual user behaviour. They often rely on opt‑in panels or API calls, which skew towards tech‑savvy users. Worse, AI responses are non‑deterministic — the same prompt produces different results because models pick words based on probability. Research by Rand Fishkin and Gumshoe.ai tested 2,961 prompts across ChatGPT, Claude and Google AI with 600 volunteers, and found less than a one‑in‑100 chance of getting the same brand list twice. In short, ranking in AI doesn't exist yet.
So why are marketers still chasing prompt volume? Partly because it feels comfortable to have numbers to track. We're used to Semrush or Ahrefs telling us how many searches a keyword gets. But we're in a pre‑Semrush era for AI search. The infrastructure to measure LLM prompts reliably doesn't exist. Tools can be useful for directional signals, but basing your entire content plan on them is like building a house on sand. Instead, the brands succeeding in AI search share a common trait: they understand their customers deeply and translate that knowledge into authoritative, helpful content.
For readers new to Infinity1, it's worth noting why this topic matters now. Many trade‑based businesses are being pitched AI "growth hacks" that promise easy wins. In reality, chasing algorithmic tricks rarely generates lasting growth. Queensland tradespeople, for example, succeed when they understand local homeowners and offer genuine advice that solves real problems. Generative search isn't some mystical force; it's another channel where authority, trust and customer insight pay off.
Why Prompt Volume Misleads Your GEO Strategy
The promise of prompt volume tools is enticing. They claim to show which AI prompts drive visibility for your brand so you can prioritise content. Yet the reality is that prompt volume is an unreliable foundation for several reasons.
Google reports search frequency because it records queries, but LLMs don't expose their data. Tools that estimate prompt volume rely on proxies or modelling. In the case of Profound, data comes from opt‑in consumer panels — but panel participants are more engaged and tech‑savvy than the general population, so their prompts don't reflect how your average customer speaks.
Traditional keyword volume works because millions of people type the same phrases into Google, producing stable metrics. AI interactions are conversational; users rephrase questions multiple times and models generate responses probabilistically. The SparkToro/Gumshoe study found that AI brand recommendations are essentially random. If the same prompt can generate entirely different lists, chasing a "rank" in AI results becomes meaningless.
Many tools simulate prompts through API calls rather than user interfaces, yet early research shows API results may diverge from human‑generated ones. Even when using human panels, citation drift is massive — Profound measured month‑to‑month variations of dozens of percentage points in citations within Google AI Overviews and ChatGPT.
Tools like Semrush or Moz took years to develop reliable SEO metrics; generative search hasn't reached that maturity. Nikki Lam notes that nobody has complete visibility into LLM impact yet and current tracking data should be treated as directional, not definitive. Building strategy around unstable and biased numbers distracts from fundamentals that actually move the needle.
Start With Your Ideal Customer (ICP)
If prompt volume isn't the answer, what is? The most reliable signal you have is your Ideal Customer Profile (ICP). Your ICP describes the people who benefit most from your product or service — their pain points, goals and language. Instead of letting a vendor's dashboard tell you what to optimise for, start with what you already know about your audience. Ask: what problems do your best customers hire you to solve? What words do they use when they're searching for solutions? Those phrases belong in your content because they reflect real intent.
For example, suppose you run a solar installation company. Instead of guessing which AI prompts might mention "top solar installers in Brisbane," talk to your customers. Why did they choose solar? Maybe they wanted to reduce energy bills, get independence from the grid or benefit from government rebates. Listen to the exact language they use — is it "solar panels," "solar system," or "PV installation"? Once you capture that language, you can create content that answers their questions directly. If you've done solid ICP work, you're sitting on better data than any prompt tool can provide.
Venture beyond your own marketing materials into forums, Reddit threads, LinkedIn comments, Slack communities and review sites like G2 and Trustpilot. In these spaces, people ask unfiltered questions in their own words, which often map closely to how they prompt AI tools. A repeated question in a subreddit is a stronger content idea than any vendor‑curated query list.
Sales calls, support tickets, onboarding sessions and interviews capture authentic questions and objections. If your team hears the same objection every week, it's a safe bet someone is asking an AI the same thing. Build content that answers those objections clearly and thoroughly — this positions you as a helpful expert in AI answers and strengthens traditional SEO.
Generative engines often draw from existing high‑ranking pages and reputable directories. Appearing in comprehensive, high‑authority list articles increases your chances of being cited. Building topical pillars, publishing new content consistently and earning backlinks remain essential to raise your site's authority. GEO doesn't replace SEO — it amplifies it.
Design an Actionable GEO Plan – Clustering, Tools & Monitoring
Once you understand your customers' language, the next step is to structure and monitor your strategy. Treat each potential prompt as part of a larger conversation rather than an isolated keyword. Cluster prompts by intent and theme so you can identify patterns in how your audience thinks about problems. For example, the overarching topic "how to measure the success of a solar installation" might include sub‑prompts about metrics (payback period, kilowatt‑hours saved), stakeholder communication (reporting ROI to the CFO) and benchmarking (comparing suppliers). Each deserves its own content piece. Clustering helps you build a coherent narrative that signals topical authority to both users and AI.
Platforms like Profound or Writesonic can be useful for spotting topic gaps, monitoring brand visibility and tracking share of voice. They can show whether your brand appears in AI answers and how you compare to competitors over time. However, avoid using these tools as a keyword volume substitute. Let your ICP, audience research and customer conversations determine what to optimise — then use prompt data to pressure‑test and monitor, not to decide.
Given the high citation drift in AI outputs, checking your brand's visibility once a quarter is insufficient. Create a defined list of 20–30 prompts that reflect your ICP's most common questions. Run them every month across platforms like ChatGPT, Perplexity and Google AI Overviews. Track whether your brand or competitors appear and note trends rather than reacting to single‑month swings. Over three to six months you'll see directional changes that can inform adjustments.
Encourage customers to leave positive reviews on industry‑specific platforms; generative engines favour verified, recent reviews from trusted sites. Monitor your social sentiment — the language people use when talking about your brand online — as it increasingly influences AI recommendations. Work on earning backlinks and publishing consistently to raise your domain authority. By combining these SEO fundamentals with ICP‑driven clustering and intelligent monitoring, you create a robust strategy that isn't thrown off by the volatility of prompt volume metrics.
Conclusion
Generative‑engine optimisation is still in its infancy. There's no magic dashboard showing the exact prompts your customers use or a stable ranking to chase. The data behind prompt volume is largely estimated and biased, and AI responses change from one interaction to the next. Studies show that brand lists in AI outputs are essentially random, and citation patterns drift wildly month‑to‑month. Relying on such shaky numbers to guide your strategy is a recipe for frustration.
The solution is to flip your approach. Put your customers at the centre. Build a detailed ideal customer profile and listen to the questions they ask in forums, reviews and your own conversations. Organise those prompts into clusters around core themes and create authoritative content that answers them thoroughly. Support this with tried‑and‑true SEO practices — publish consistently, earn backlinks, secure placements in authoritative lists and directories, encourage reviews and monitor social sentiment. Use prompt‑tracking tools for directional awareness, but let your ICP guide your priorities.
By adopting this customer‑first mindset, you'll build topical authority and trust that resonates not only with generative engines but also with human readers. Over time, monthly monitoring of your key prompt clusters will reveal meaningful trends. Instead of chasing an elusive "prompt volume," you'll invest in content that addresses real pain points and positions your brand as the expert.
At Infinity1 we've seen this play out across dozens of campaigns. When we pivoted a client away from generic AI prompts toward answering the specific questions their customers asked on forums and in sales calls, their organic traffic and leads improved — not because a tool predicted it, but because people found value in their content. That's the essence of GEO: aligning what you produce with what people actually need, then giving AI models something credible to cite. If you'd like help building a customer‑driven GEO plan, reach out. We love collaborating with businesses that care about serving their customers.
Frequently Asked Questions
What is Generative Engine Optimisation (GEO)?
Generative Engine Optimisation (GEO) is the practice of optimising your business, content and digital presence so that AI-powered tools — like ChatGPT, Perplexity and Google AI Overviews — recommend your brand in their generated answers. Unlike traditional SEO (which focuses on ranking in a list of links), GEO focuses on becoming the cited authority inside AI-generated responses.
How is GEO different from traditional SEO?
Traditional SEO optimises for ranking in a list of ten blue links on a search results page. GEO optimises for being the answer that an AI model generates in natural language. GEO relies on entity authority, structured content, topical depth, review signals and factual accuracy rather than just keywords and backlinks. Both remain important — GEO doesn't replace SEO, it builds on top of it.
Should I try to target specific AI prompt keywords?
Not as your primary strategy. AI responses are non-deterministic — the same prompt produces different outputs because models pick words based on probability. Tools claiming to measure AI prompt volume are estimating, not reporting real data. A more durable strategy is to build deep understanding of your customers' real questions (from forums, sales calls, reviews) and create authoritative content that answers those questions thoroughly.
How do I measure GEO success?
Track your brand's presence across AI tools by running a defined set of 20–30 customer-relevant prompts monthly across ChatGPT, Perplexity and Google AI Overviews. Note whether your brand is cited, and track trends over three to six months rather than reacting to single-month results. Also monitor referral traffic from AI tools in Google Analytics and track changes in brand search volume over time.
Explore Infinity1 services:
Want Infinity1 to help your business?
We specialise in AI-driven marketing for tradespeople and eCommerce businesses. Get in touch for a free consultation.
Contact Infinity1