SEO »
You can't measure every AI recommendation, but you can build a reliable picture of your brand's presence. Here's how. #
AI search doesn’t work like traditional search. A prospect might ask ChatGPT for the best CRM for manufacturing companies, compare options in Google’s AI Mode, refine their requirements over several follow-up questions, and make a shortlist — all without ever clicking a website.
If your company appears in those conversations, you’ve influenced the buying process. The challenge is proving it. Prompt-level visibility measurement has become one of the fastest-growing areas of AI search optimization. It’s also one of the most misunderstood. Many vendors promise complete visibility into AI conversations, but the reality is far messier.
Here’s what you can measure today, what you can’t, and how practitioners are building useful reporting despite the current limitations.
A 5-step framework for tracking AI visibility #
1. Accept that AI doesn’t have ‘rankings’
The biggest mistake marketers make is trying to recreate traditional SEO reports — because there is no universal “position 1” inside ChatGPT.
The same prompt may produce different responses based on:
- Conversation history
- User location
- Personalization
- Follow-up questions
- Model version
- Available web retrieval
- Time
Visibility is now probabilistic rather than deterministic. Instead of asking, “Do we rank?” the better question is, “How often are we included across the conversations that matter?”
That shift changes everything about measurement.
[ Be the brand AI recommends. See your AI visibility
](https://www.semrush.com/ai-seo/overview?utm_campaign=ic_sel_0101ai&utm_source=searchengineland.com&utm_medium=overlay&onboarding=off) See where your brand appears in AI search, where competitors are winning, and what it takes to become the answer AI recommends.
2. Build a prompt library instead of a keyword list
Keywords are still useful. They’re just no longer enough.
Instead of tracking individual search terms, build a library of prompts that reflect how real people research purchases.
The easiest way is to organize prompts by search intent:
Intent | Example prompt | | Discovery | What are the best workforce management platforms? | | Comparison | Rippling vs BambooHR vs Deel | | Evaluation | Which HR platform works best for global hiring? | | Validation | Is Company X actually worth the cost? | | Objections | What are the disadvantages of using Company X? | | Alternatives | What should I use instead of Company X? | | Implementation | How difficult is Company X to implement? |
Instead of monitoring 10 keywords, you might monitor 200 to 500 prompts covering the entire buying journey. That produces a much more realistic picture of AI visibility.
3. Use prompt clusters, not individual questions
One prompt rarely tells you anything useful.
For example, “best CRM software” might not mention your company. But “best CRM for manufacturing companies” might. And “CRM for manufacturers with field sales teams” might produce completely different recommendations.
Instead of focusing on individual prompts, group [similar prompts into clusters](https://searchengineland.com/prompt-research-seo-geo-strategy-471399).
For example:
Category cluster- Best project management software.
- Best PM platform.
- Project management tools.
Industry cluster- Best CRM for healthcare.
- Best CRM for manufacturing.
- Best CRM for finance.
Feature cluster- CRM with AI automation.
- CRM with forecasting.
- CRM for enterprise sales.
Patterns across clusters are far more reliable than individual prompt results.
4. Mix synthetic prompts with real user questions
This is where measurement gets tricky.
Most organizations don’t know what customers are actually typing into AI assistants. So they generate prompts synthetically.
That usually involves:
- Expanding keyword researchinto conversational questions. - Generating prompt variations with AI.
- Creating comparison, objection, and follow-up prompts.
Synthetic prompts are valuable because they’re repeatable. But they have limitations. Generated prompts often sound cleaner and more structured than real user behavior.
Actual conversations tend to look more like:
- “We’re a 250-person SaaS company with a small HR team. We already use Workday but need something better for payroll. Budget isn’t a huge issue. What would you recommend?”
That’s much richer than “best payroll software.”
The best measurement programs use synthetic prompts for consistent benchmarking, then supplement them with real prompts collected from sources like:
- Sales calls.
- Customer interviews.
- Support conversations.
- Community discussions.
- Internal search logs.
- On-site search.
- AI transcripts that customers voluntarily share.
No prompt library stays accurate forever. It should evolve as customer language changes.
5. Measure multi-turn conversations #
Most AI buying journeys don’t happen in a single prompt. Someone might start by asking for the best cybersecurity vendors, then narrow the list to those strongest for healthcare, ask which ones integrate with CrowdStrike, and finally compare pricing.
Your company may not appear in the first response. But it might become highly recommended by the third.
If you’re only measuring the opening prompt, you’ll miss a large portion of meaningful visibility. Modern prompt tracking should evaluate entire conversation paths, not just isolated questions. That often reveals different patterns than single-shot testing.
Metrics that actually matter #
Many traditional SEO metrics don’t translate neatly to AI search. Rankings, clicks, and impressions still have value, but they no longer tell the whole story.
Instead, marketers are beginning to rely on different measurements that better reflect how brands appear — and how they’re positioned — inside AI-generated responses.
Inclusion rate
If you only track one AI visibility metric, make it this one. Inclusion rate measures the percentage of tracked prompts where your brand appears in the AI’s response. For example, if you monitor 500 prompts and your company is mentioned in 185 of them, your inclusion rate is 37%.
On its own, that’s a useful benchmark. It becomes even more valuable when you segment it by factors like buying stage, product category, industry, geography, or AI model. Those slices often reveal opportunities that an overall average would hide.
Position within the response
Being mentioned isn’t the same as being recommended.
It’s worth tracking whether your brand is the first recommendation, one of the first few options, buried near the end of a list, or mentioned only as an alternative. If the response includes a comparison table, note where your company appears there as well.
AI answers don’t have traditional rankings, but prominence still matters. A top recommendation is naturally more likely to shape a buyer’s perception than a passing mention several paragraphs later.
Brand framing
Visibility tells you whether you’re included. Brand framing tells you how you’re being described.
For example, there’s a meaningful difference between an AI describing your company as “widely considered an enterprise leader” and “best suited for smaller teams.” Both are positive, but they position the brand very differently. Look for recurring themes around strengths, weaknesses, differentiators, pricing, ideal customer profile, and competitive comparisons. Over time, these patterns can highlight messaging gaps in your own content — or reveal how the broader web is shaping AI’s understanding of your brand.
Sentiment
Sentiment goes beyond simply labeling responses as positive or negative. It also captures the confidence with which AI presents your brand.
Compare these two statements:
- “Company A is generally considered the strongest option…”
- “Company A may be worth considering.”
Neither is negative, but they carry different levels of conviction. Tracking confidence, uncertainty, caution, skepticism, and strong endorsement can provide a more nuanced view of how AI systems are presenting your company to prospective buyers.
Competitive share of voice
Your own visibility is only part of the picture. It’s equally important to understand how often competitors appear alongside — or instead of — you.
For example, if your inclusion rate stays at 40% month after month, that might seem disappointing. But if every major competitor also dropped by 20 percentage points after a model update, the story changes. Conversely, if one competitor suddenly jumps from appearing in 35% of prompts to 70% while everyone else remains flat, that’s worth investigating.
Competitive share of voice helps you distinguish category-wide shifts from changes specific to your brand.
The current tool landscape #
Over the past year, the market for AI visibility platforms has expanded quickly. While each product approaches the problem a little differently, most are trying to answer the same core questions:
- Does my brand appear?
- How often?
- In which AI models?
- Against which competitors?
- And how is it being described?
Many platforms also include features like prompt libraries, competitive benchmarking, citation tracking, answer monitoring, and trend reporting. These capabilities can dramatically reduce the manual effort required to test hundreds or even thousands of prompts on a recurring basis.
That said, it’s important to understand what these tools are – and aren’t – measuring.
None of them has access to every AI conversation happening in the wild. Most rely on controlled prompt libraries, repeatable testing environments, or sampled interactions to create a representative view of visibility.
That’s incredibly useful, but it isn’t the same thing as observing every real user interaction.
What you still can’t reliably track #
This is the part many vendors tend to gloss over.
Despite how quickly AI measurement has evolved, there are still important things that simply aren’t observable today. No platform can comprehensively tell you:
- Every prompt where your brand appeared.
- Every conversation that influenced a purchase.
- Every recommendation made inside ChatGPT.
- Every citation shown to every individual user.
- Exactly how personalization changed a response.
- Every multi-turn conversation across every AI platform.
- How often someone acted on an AI recommendation without ever clicking a link.
The underlying AI platforms simply don’t expose that level of data. If a vendor claims they can see every AI conversation involving your brand, it’s worth asking exactly how they’re collecting that information.
What a practical measurement framework looks like #
Rather than chasing perfect attribution, the goal should be to build a repeatable measurement system that you can track consistently over time.
A practical dashboard might include four categories of metrics:
Visibility
Inclusion rate.- Competitive share of voice.
- Prompt coverage.
- Model coverage.
Response quality
- Position within the response.
- Brand framing.
- Sentiment.
- Message consistency.
Technical signals
- Citation frequency.
- Content retrieval success.
- Entity consistency.
- Freshness.
Business outcomes
AI referral traffic.- Assisted conversions.
- Branded search lift.
- Direct traffic trends.
- Pipeline influenced by AI discovery.
No single metric tells the whole story. Together, however, they provide a much more complete picture of how your brand is showing up — and being perceived — across AI-assisted research.
[
If AI can’t find you, customers won’t either.
See your AI visibility
](https://www.semrush.com/ai-seo/overview?utm_campaign=ic_sel_0102ai&utm_source=searchengineland.com&utm_medium=overlay&onboarding=off) Track your visibility across AI search, uncover missed opportunities, and grow your presence where customers are asking questions.
The goal isn’t perfect measurement #
Prompt-level visibility isn’t as mature today as keyword tracking became over the past two decades.
Some signals are still emerging, others remain inaccessible because AI platforms don’t expose the underlying data, and user behavior continues to evolve almost as quickly as the technology itself.
That doesn’t mean measurement is impossible. It simply means the objective has changed.
Instead of trying to reconstruct every AI conversation, focus on:
- Building a representative prompt library.
- Tracking visibility consistently over time.
- Benchmarking against competitors.
- Understanding
how your brand is being framed. Those trends are far more actionable than chasing a level of precision the ecosystem can’t yet support.
The organizations making the most progress in AI search aren’t waiting for perfect attribution. They’re establishing consistent baselines, watching for meaningful movement, and adapting as both the models and user behavior continue to evolve.
Contributing authors are invited to create content for Search Engine Land and are chosen for their expertise and contribution to the search community. Our contributors work under the oversight of the editorial staff and contributions are checked for quality and relevance to our readers. Search Engine Land is owned by Semrush. Contributor was not asked to make any direct or indirect mentions of Semrush. The opinions they express are their own.