If you want to measure and optimize your brand’s visibility in AI platforms, you’ll need to generate a set of sample prompts for your AI visibility tool to track. And the quality of the reports you get from any tool depend on how strategically you’ve built that sample prompt set.
Good news! We’ve put together a list of principles to keep in mind when building that sample prompt set:
- Focus on topics that connect to actions and results
- Reverse-engineer the AEO reports you want
- Get stakeholder input and buy-in on your prompt set
- Inform your prompt set with existing marketing sources
- Build your prompt set pre-tool
- Write prompts from your audience’s perspective
- Build your prompt set with competitor parity
- Use platform-specific prompts (not one-size-fits-all)
- Give each report enough prompts to yield meaningful numbers
- If you’re on a budget, prioritize neutral prompts
And of course, if you want a hand doing this at your company, we’d love to explore how we can help.
Let’s go!
Why AEO prompt architecture is important
If you don’t set up your sample prompts strategically, you’ll waste a lot of time and a lot of money to get a lot of noise.
AEO report quality depends on strategic prompts
Most of the major answer engine optimization tools (and all the best ones today) rely on you creating a sample set of prompts. These tools feed those prompts into AI platforms over and over again, and bring the response data back to you. Any metrics regarding share of voice, mention rankings, sentiment, and citations that these tools give you comes from the responses to the prompts you tell it to input.
That means even if you’re using one of the best options on the market today, the reports you get back will only be as useful as the prompts you put in. If you’re feeding your AI visibility tool prompts that don’t strategically map to your desired outcomes (or worse: letting them generate your prompts for you with AI), the numbers you get back won’t help your marketing strategy, either.
Prompt strategy mitigates expenses
Every major AEO tool’s plans come with a predetermined prompt cap, and you pay more to add extra prompts to the mix. Those initial prompt limits may seem like plenty at the beginning, but they burn up fast.
| Brand | Starter plan | Mid-tier plan |
| Profound | 50 prompts $99/mo | 100 prompts $399/mo |
| Scrunch | 350 prompts $300/mo | 700 prompts $500/mo |
| Peec | 25 prompts €89/mo (~$105 USD) | 100 prompts €199/mo (~$235 USD) |
| Rankscale | 480 responses $20/mo | 4800 responses $99/mo |
Plus, some tools will charge you per prompt per platform as well. For example, if you want to track the prompt “What are the top three enterprise sales enablement software platforms?” on Perplexity, ChatGPT, and Google AI Overviews you could be looking at:
1 prompt text x 3 platforms = 3 prompts
So for example, Scrunch’s starter plan includes 350 prompts. However, every individual prompt is assigned a single platform. Scrunch currently tracks across seven platforms, which means if you were to track the prompt “What are the best alternatives to My Brand?” across all of them, that single input would cost you seven prompts—bringing you down to only 50 text inputs.
Prompts are precious when it comes to AEO tool usage, so you need to make sure you’re only using ones that make sense. A well thought-out prompt architecture helps you do this, ensuring you’re getting as much value as you can from your plan.
A predetermined prompt strategy enables smarter tracking over time
Most of these tools only begin recording AI responses to prompts when you start tracking them. They’re not giving you access to the vault of previous responses to identical prompts tracked by other marketers.
This has a few implications when it comes to introducing new prompts to your reports.
When you add, change, or reduce your prompt set, the tool will stop tracking the old prompts and begin tracking the new ones. Your month-over-month reports will no longer be apples-to-apples comparisons—which means you’ll need to manually establish baselines for any before and after reports.
Some of this is unavoidable. You can’t expect your prompt set to remain static over the years. But if you spend your first year on the tool tweaking your prompt set until you’re happy with it, you won’t get a clear idea of how your efforts are impacting your AI visibility.
A strong prompt architecture helps you determine what prompts you need to track ahead of time, so you can start tracking results after your first month.
10 principles for building a strategic prompt set
There is no cookie-cutter template for building a sample prompt set. Your AEO prompt architecture should be specific to your own marketing strategy. However, there are a few universal principles that you should keep in mind when building a prompt set.
1. Focus on topics that connect to actions and results
Your AEO tool will run any prompt you feed it, and then tell you what the target LLM had to say in response at that moment. Some of these tools will provide you with AI-generated advice for next steps (and it’s not always useful—or even sensical). It’s on you to turn what your AEO tool gives you into something you can work with.
Ideally, your prompt architecture should set up your AI visibility tool to bring you information that falls into at least one of these two categories:
- Stuff you can do something about
- Stuff that shows how what you’ve already done is working
If your AEO reports give you information that doesn’t fall into at least one of these categories, you’re wasting time and labor.
Therefore, limit your prompt set to bring back responses that you can materially work with. For most SMBs, this will likely mean using prompts that directly reference your offerings, your competitors’ offerings, the categories your offerings occupy, and any specific marketing awareness efforts you’ve made.
2. Reverse-engineer the AEO reports you want
You’ll be using your AI visibility tool to generate reports on your brand’s performance, so the first thing you should do is determine what information those reports will include. These reports should be informed by your marketing strategy’s approach to market segmentation as well as the campaigns and initiatives in your marketing strategy.
Create a draft of the kind of reports that you’d want to present to both leadership and the rest of your team. What are the important metrics that they need to see? What specific topics and issues are they most concerned with in regards to AI responses? What information will be most helpful to send to your content, PR, product, customer support, and SEO teams?
Your sample prompt set should set up your AI visibility tool to produce these answers.
3. Get stakeholder input and buy-in on your prompt set
Since you’ll be sharing these AEO reports with other people in your organization, you’ll want to get their input on the kinds of topics they’ll want to see data on. Depending on your budget for AI visibility tools, you might want to include a prioritization exercise in this process. This will help you develop an understanding of which topics or specific prompts the people reading your reports are most interested in.
Beyond this, you should document the rationale behind your prompt architecture so that every stakeholder can quickly understand exactly what prompts are included in a given report. Not only is this courteous, but it also protects you when someone in the organization says, “I asked ChatGPT what the best software for sales professionals is, and it didn’t mention us at all!”
4. Inform your prompt set with existing marketing sources
Some of the prompts you choose will likely be speculative—but you should be able to trace the majority of your prompt choices to actual marketing sources. This includes not only the topics that your prompt set covers, but also the wording used in those prompts.
Some of these sources are internal. Examples include:
- High-performing organic search keywords and topics
- High-performing SEM keywords and topics
- Key language from sales materials
- High-traffic customer support pages
- User chatbot analytics
- Product usage data
But you should also look to external sources for inspiration. Browse pages on review aggregator sites like G2 and Capterra to identify generic key value drivers. Analyze your competitors’ product pages to see what appeals they’re making to the people you want to sell to. Follow influencers on LinkedIn and determine which topics are generating interest in your market.
Doing this will ground your AEO prompt set in the topics that real people have demonstrated interest in.
5. Build your prompt set BEFORE getting an AEO tool
You’ll be tempted to get your AEO tool first and then build your prompt set within that tool—but that comes with serious drawbacks. While you can strategically build your prompt set in-tool, you’ll have a much easier time doing so in a spreadsheet first. There are a few reasons this is the case.
For starters, both your organization and AI visibility tools want you to get set up and reporting as quickly as possible. Every day you’re paying for a tool and not using it is wasted cash on your end, and increased churn potential on the vendor’s.
But architecting a strategic prompt set takes time, consideration, and collaboration. That means that if you start your tool subscription before designing your prompt set, you’ll face the following options:
- Pay for a tool you’re not using while you figure out your prompt set
- Quickly develop a prompt set using the tool and manually adjust your reporting to account for the inevitable changes you will make
- Quickly develop a prompt set to start using the tool and don’t make any corrections to reporting
All of these waste money. The first wastes money on the tool itself. The second wastes money on the labor that it takes to form any before/after comparisons. And the third wastes money on both tool expenses and reporting labor—because your reports won’t accurately reflect your marketing activities.
You’re much better off building your prompt set before you start paying for the tool. You’ll avoid paying for a tool before you can get value from it, and you’ll also have a resource that any relevant stakeholders in your organization can reference to view and understand your prompt set—without needing to pay for extra seats on your AI visibility tool of choice.
6. Write prompts from your audience’s perspective
For skilled marketers, much of this goes without saying:
- Your prompts should imitate the natural prompts that people would use for these platforms. Your prompts should be written from the answer engine user’s perspective, not your own brand’s.
- Your prompts should be about topics your audience will reasonably be curious about. (You already have a basic understanding of this, but you can use Sparktoro to expand on this.)
- Your prompts should use language your audience actually uses. (Referencing sales call recordings, Google’s “People Also Ask” boxes in SERPs, and the “Questions” tool in Ahrefs’ Keywords Explorer are easy places to start here.)
- Your prompts should include parameters and qualifiers that your ideal customer profile would reasonably use. (Your internal strategy docs, as well as review aggregator sites like G2 can be helpful for this.)
One thing that might not be obvious on this front is that every single AI-generated prompt should pass a careful human review. Almost every tool will suggest prompts related to the websites, product categories, and topics you load up into it. However, these prompts are (at best) helpful for covering very basic topics and idea generation. You know your audience better than these bots—don’t let them eat up your limited prompts by accepting their vanilla suggestions.
7. Build your prompt set with competitor parity
If you want a thorough view of how you’re showing up in answer engines, you need to understand how your competitors are showing up as well. At the very least, this will allow you to compare the way LLMs treat your brand to the ways they treat your rivals.
More importantly, it can save you from wild-goose-chase labor expenses. For example, let’s say one of your tracked prompts is:
“What are the pros and cons of using My Brand for enterprise CPG companies?”
By asking for cons, the answer engines will necessarily bring up a list of problems with your offerings—and your AEO tool will advise you to create content (or even adjust your product) to address these “grievances.” If you’re selling a B2B SaaS product, those grievances will almost always reference price and learning curve. If you’re only looking at what LLMs say about your brand, this might signal that your pricing and user friendliness are major problems that need to be addressed.
But now imagine that you tracked all of these prompts:
“What are the pros and cons of using My Brand for enterprise CPG companies?”
“What are the pros and cons of using Competitor A for enterprise CPG companies?”
“What are the pros and cons of using Competitor B for enterprise CPG companies?”
“What are the pros and cons of using Competitor C for enterprise CPG companies?”
Now you can see not only the pros and cons listed for your brand, but also the pros and cons listed for each of those competitors. If your AEO tool has strong sentiment analysis capabilities (like Rankscale—they are the best at this), you can see how LLMs treat every player, and with an extra layer of manual work, you can determine which pros and cons are generic (i.e., LLMs say this about everyone), which ones are specific to you, and which ones are specific to competitors.
What would have looked like a serious problem for your brand might turn out to be something answer engines say about all your competitors—which means it’s probably not something your marketing, sales, product, and/or executive teams need to worry about.
8. Make platform-specific prompts
Every major AI visibility tool will allow you to enter prompts and assign them to as many answer engine platforms as they can. This saves you the time of entering the same prompt over and over for different platforms.
But it also means you’ll be entering prompts into platforms that don’t make sense.
Think about how you use Google vs. how you use ChatGPT. It’s entirely reasonable that you would type “software component analysis tools” into Google—but is that really how you would prompt ChatGPT to give you a list of the best SCA tools?
This is the kind of thinking you need to bring to your prompt architecture.
The prompts you tell your AI visibility tool to track in Google AI Overviews should not be the same as the ones you tell it to track in ChatGPT. Your prompt architecture should account for the various ways people use these answer engines, instead of crop-dusting all the platforms with all the prompts.
9. Give each report enough prompts to yield meaningful numbers
If you’re following the advice in principle #1, your prompt set will be built to provide you with metrics for every report you anticipate building. If you’re doing this well, then each potential report will include enough prompts to provide numbers that are actually useful to you.
Your AEO tool–generated reports will be based on the prompts you feed those tools. This means that your metrics (mentions, citations, sentiment, etc.) will all come down to the things you’ve specifically told your tool to enter into relevant platforms. These tools will usually run those prompts on a daily basis and give you a real-time report on what they get back from answer engines.
That means that if you only have two prompts for a given report, then your mention outputs on any given day will be either 0%, 50%, or 100%. That’s not especially helpful when it comes to measuring trends over time (especially since we know these platforms aren’t consistent in their responses anyway). You need a decent cluster of prompts per report in order to get meaningful numbers—otherwise your metrics will fly between “nothing” and “everything.”
So, how many prompts should you use per report? Rankscale co-founder Patrick Schmid told me in an interview that they recommend using 7–10 prompts per topical report. This gives you a decent view of how the topic is treated by LLMs without burning up an ungodly amount of your prompt budget.
10. If you’re on a budget, prioritize neutral prompts
LLMs are fantastic at telling you what you want to hear. If you tell ChatGPT to give you a list of 20 reasons an enterprise client should use your offerings, it will deliver. If you tell Perplexity to give you 20 reasons an enterprise client should not use your tool, it will deliver. When you’re dealing with AI answer engines, every question can be a leading question.
So if you’re working on a budget, your sample prompt set will cover a lot more ground by simply addressing neutral questions.
For example, the prompt:
“What are the top 5 [my product category] solutions for [a key problem you solve]?”
Will give you a list of players that LLMs consider candidates. You’ll know if (and where) you place in their responses. This is a far more economical way of learning how these platforms treat your brand than seeding them with leading questions.
Does it give you the specificity that more direct prompts about your brand would? Absolutely not. But if you’re trying to stay within an approved AEO tool budget, it gives you useful information without burning through dozens (or even hundreds) of prompts.
Bottom line: Build your prompts strategically
More than anything else, your sample prompts will determine the value you ultimately derive from your AEO tools. If you don’t approach this strategically, you’ll be creating a lot more reporting (and potentially content production) work for yourself that in’t attached to your marketing strategy.
It sounds like a tall order—because it is.
This guide gives you a general direction when it comes to setting up your AEO prompt architecture, but if you’d like to build a prompt set specific to your organization, drop us a line.
