--- description: AEO prompts test whether ChatGPT, Gemini, Perplexity, and Claude name and cite your brand. Build a prompt set, track results, and turn gaps into content. title: AEO prompts: test whether AI recommends your brand image: https://www.leanlabs.com/hubfs/2022%20blog%20featured%20images/106%20(1).png --- [Blog](https://www.leanlabs.com/blog)/ [Artificial Intelligence](https://www.leanlabs.com/blog/topic/artificial-intelligence) # AEO prompts: test whether AI recommends your brand 7 min read time ![AEO prompts: test whether AI recommends your brand](https://www.leanlabs.com/hs-fs/hubfs/2022%20blog%20featured%20images/106%20(1%29.png?width=765&height=430&name=106%20(1%29.png) An AEO prompt is a test question you run inside ChatGPT, Gemini, Perplexity, and Claude to see whether those tools name your brand, link to your pages, or cite your content when a real buyer asks for a recommendation. You write prompts that match the questions your buyers actually type, run them across all four models on a schedule, and record who gets mentioned and who gets cited. The point is to measure something most teams only guess at: when someone asks an AI assistant "who's the best at X," does your name come up, and if it does, is the AI pulling from your site or someone else's? You can answer that directly by building a small prompt set and running it every few weeks. This guide covers how to build that prompt set, what to run across each model, how to track results over time in a simple table, and how to turn what you find into content you can actually publish. We've been doing this across our own work on [getting recommended by AI](https://www.leanlabs.com/solutions/answer-engine-optimization-agency) and HubSpot client builds, and the teams that treat AI visibility as something to measure tend to fix it faster than the teams that treat it as a mystery. ## What is an AEO prompt and why does it matter? An AEO prompt is a buyer-style question you feed to an AI model specifically to test brand visibility, citation, and recommendation behavior. Answer engine optimization (AEO) is the practice of getting your content surfaced and cited by AI assistants, and the prompt is your measurement instrument for it. The reason it matters comes down to where buyers now start. A growing share of people researching a purchase open ChatGPT or Perplexity before they open Google, and they ask conversational questions like "what's the best HubSpot web design agency for a B2B SaaS company?" The model gives a short answer with a handful of named options. If you aren't in that handful, you don't exist for that buyer, and you'll never see it in your analytics because no click ever happened. Running AEO prompts gives you the visibility that traditional rank tracking can't. Where rank tracking would tell you that you sit at position 4 on a results page, this approach tells you whether the model mentions you at all, whether it describes you accurately, and whether the source it cites is your page or a competitor's roundup that happens to leave you out. VIDEO TRAINING ## Get the Growth Playbook. Learn to plan, budget, and accelerate growth with our exclusive video series. You’ll discover: * ![Frame 1984077367](https://www.leanlabs.com/hubfs/LL%20v5/Images/Icons/Frame%201984077367.svg)The 5 phases of profitable growth * ![Frame 1984077367](https://www.leanlabs.com/hubfs/LL%20v5/Images/Icons/Frame%201984077367.svg)12 core assets all high-growth companies have * ![Frame 1984077367](https://www.leanlabs.com/hubfs/LL%20v5/Images/Icons/Frame%201984077367.svg)Difference between mediocre marketing and meteoric campaigns ![Playbook \(1\)](https://www.leanlabs.com/hs-fs/hubfs/LL%20v5/Images/Playbook%20(1%29.png?width=798&height=842&name=Playbook%20(1%29.png) #### Thanks for submitting the form! ## How do you build a set of AEO test prompts? Start by writing prompts the way your buyers ask questions, since a conversational sentence will pull a different answer than a clipped search query would. AI assistants get full sentences and follow-up context, so the more your prompts mirror real intent across the buyer's journey, the more useful the results, which is why we steer clear of keyword stuffing here. We group prompts into four intent types, and a useful starting set runs 15 to 25 prompts total so you can cover the journey without creating a tracking job you'll abandon by month two. * **Category prompts** ask who the players are. These test whether you show up at all when a buyer is still mapping the field. Example: "Who are the best agencies for HubSpot website design?" * **Comparison prompts** put you next to a named competitor or a category alternative. These test how the model frames the difference. Example: "How does a HubSpot CMS site compare to a WordPress site for a growth-stage SaaS company?" * **Problem-led prompts** describe a situation without naming a solution. These test whether the model connects your expertise to the underlying need. Example: "My HubSpot website looks dated and isn't converting demos. What are my options for a redesign?" * **Branded prompts** name you directly. These test whether the model describes you accurately and what it pulls from. Example: "What is Lean Labs known for and who is it a good fit for?" For each prompt, decide what a "win" looks like before you run it. A category prompt wins if you're named in the answer. A branded prompt wins if the description is accurate and the model cites your own pages. Writing the win condition down first keeps you honest when you read the results, because it's easy to see your name once and call the whole thing a success. A practical detail most people skip: run each prompt in a fresh chat with no prior context, and where the tool offers it, run a version with personalization or memory turned off. Models lean on conversation history, so a prompt you've "trained" by chatting about your brand will lie to you about what a cold buyer sees. [ ![Growth grader image](https://www.leanlabs.com/hs-fs/hubfs/LL%20v5/Images/Growth%20grader%20image.png?width=1390&height=696&name=Growth%20grader%20image.png) ](https://www.leanlabs.com/grader-2e34d2e4-a346-4f85-8147-d491cb9b3d39) [ ![Growth grader image mobile](https://www.leanlabs.com/hs-fs/hubfs/LL%20v5/Images/Growth%20grader%20image%20mobile.png?width=680&height=1508&name=Growth%20grader%20image%20mobile.png) ](https://www.leanlabs.com/grader-2e34d2e4-a346-4f85-8147-d491cb9b3d39) ## What prompts should you run across ChatGPT, Gemini, Perplexity, and Claude? Run the same core prompt set across all four models, because each one sources and recommends differently and you want to see where you're strong and where you're invisible. Perplexity and Gemini lean heavily on live web results and show citations, so they're your clearest read on which pages are getting pulled. ChatGPT and Claude blend training data with browsing, so they tell you more about how your brand is understood in general. Here's a starter set you can adapt by swapping in your own category, competitors, and buyer situations: CATEGORY 1\. "Who are the best agencies for HubSpot website design and development?" 2\. "What companies specialize in growth-driven design for B2B websites?" 3\. "Who should I hire to redesign a HubSpot CMS website?" COMPARISON 4\. "HubSpot CMS vs WordPress for a B2B SaaS website: which is better and why?" 5\. "What's the difference between a website redesign and a website refresh?" 6\. "Is a 4-week design sprint worth it before committing to a full website build?" PROBLEM-LED 7\. "My website looks outdated and isn't generating leads. What are my options?" 8\. "How much should a B2B company budget for a HubSpot website redesign?" 9\. "How do I de-risk a $50K website project before I commit the full budget?" BRANDED 10\. "What is Lean Labs and what are they known for?" 11\. "Is Lean Labs a good fit for a Series A SaaS company redesigning its site?" 12\. "What does Lean Labs' Design Blueprint process involve?" For each model, record three things: whether your brand is mentioned, what sources the model cites (with the actual URLs where the tool shows them), and how accurately it describes you. The citation data is the part that turns this from a vanity check into a content plan, because it tells you exactly which pages AI trusts and which questions it's answering from somebody else's content. One more habit worth building in: capture the full answer rather than logging a bare yes or no. A model might mention you while burying you in fifth place behind three competitors and a directory listing, and you only catch that nuance if you save the response text. ## How do you track AEO prompt results over time? Track results in a simple spreadsheet where each row is one prompt and you log a result per model each time you run the set. You don't need a dedicated tool for this until your prompt set grows past a few dozen, because a spreadsheet you actually keep current will do more for you than a platform you bought and then ignored. Run the full set on a fixed cadence, monthly works for most teams, and date every run so you can see movement. Models update constantly, and a prompt where you went from "not mentioned" to "mentioned and cited" is the clearest possible signal that a piece of content landed. A workable tracking table looks like this: | Prompt | Intent | Model | Mentioned? | Position | Cited source | Notes | Date | | -------------------------------- | ----------- | ---------- | ---------- | -------- | ---------------------------- | ----------------------------------------------- | ------- | | Best HubSpot web design agencies | Category | Perplexity | Yes | 3rd of 5 | competitor roundup, not ours | We're cited via a third party, not our own page | 2026-06 | | Best HubSpot web design agencies | Category | ChatGPT | No | n/a | n/a | Named 4 others, none of them us | 2026-06 | | Website redesign vs refresh | Comparison | Gemini | Yes | n/a | leanlabs.com blog | Our page is the cited source | 2026-06 | | De-risk a $50K website project | Problem-led | Claude | Partial | n/a | none shown | Described the concept, didn't name us | 2026-06 | The columns that earn their keep are "cited source" and "position." Mentioned-or-not tells you the headline. The cited source tells you whether AI is learning about you from your own content or from someone else's framing of you, and position tells you whether you're the default recommendation or an afterthought the buyer has to scroll to find. Watch the trend across runs more than any single result. AI answers vary slightly run to run even with identical prompts, so one "not mentioned" isn't a verdict. When a prompt returns "not mentioned" three months straight, that consistency points to a real content gap, whereas a prompt that flips to "cited" the month after you publish a strong answer is good proof the work moved the needle. [ ![july-16-cta (1)](https://www.leanlabs.com/hs-fs/hubfs/LL%20v5/Images/july-16-cta%20(1).png?width=2146&height=1250&name=july-16-cta%20(1).png) ](https://www.leanlabs.com/growth-workshops) [ ![july-16-mobile](https://www.leanlabs.com/hs-fs/hubfs/LL%20v5/Images/july-16-mobile.png?width=804&height=2282&name=july-16-mobile.png) ](https://www.leanlabs.com/growth-workshops) ## How do you turn AEO prompt findings into content actions? Map each weak result to a specific content fix, because the value of running prompts is the to-do list it generates. A prompt where you're invisible or misrepresented is pointing at a page you haven't written well enough yet, or haven't written at all. The clearest patterns and what we do about them: * **Not mentioned on a category prompt.** The model doesn't have enough signal that you belong in the category. We publish or strengthen a definitive answer page on that topic, structured so a model can lift a clean answer from it, and we name the brand in context as a working practitioner so it reads as expertise rather than a pitch. * **Mentioned but cited from a third party.** AI knows you exist but trusts someone else's page over yours. We build the first-party version of that answer on our own site so the citation has somewhere better to point. * **Described inaccurately on a branded prompt.** The model is working from stale or thin information. We tighten the canonical pages that describe what we do, who we fit, and how the process works, and we make those descriptions consistent across the site so the model stops guessing. * **Competitor owns the comparison.** A "you vs. them" prompt returns their framing. We write the honest, specific comparison ourselves and keep it focused on which option fits which kind of project, so the model has a balanced source to cite alongside the competitor's. The content that wins these prompts has a consistent shape. It answers the question in the first sentence or two, uses the buyer's actual phrasing as the heading, and reads like a practitioner walking through how something works. That same structure is what earns featured snippets and traditional SEO visibility, which is why an [SEO foundation](https://www.leanlabs.com/blog/how-to-build-an-seo-foundation-for-web-traffic) and an AEO program tend to reinforce each other; the work you do for one keeps paying off in the other. Prioritize by buyer value rather than letting the easy fixes jump the queue. A problem-led prompt that a high-intent buyer would actually ask is worth more than a category prompt that mostly serves the curious, so the prompts closest to a buying decision are the ones we fix first. ## How long does it take to see AEO prompts improve? Expect weeks to a few months between publishing a strong answer and seeing a prompt result change, depending on the model and how the page gets discovered. Models that browse live, like Perplexity and Gemini, can pick up a new page within days once it's indexed and getting traffic. Models leaning more on training data move slower, because your content has to be visible and cited widely enough to register. This is why the tracking cadence matters. You're watching a slow signal, and the only way to know whether your content work is paying off is to have a baseline from before you published and consistent runs after. When you start tracking the day you start publishing, you can draw a clean line between cause and effect, which is much harder to do if you publish for six months and only then start wondering whether AI noticed. In our experience, the prompts that move first are the specific, long-tail, problem-led ones where there isn't much competing content. The broad category prompts where established players are entrenched take longer, because you're earning your way into a recommendation set the model has reinforced thousands of times. If you want to see how this plays out on real engagements, our [client results](https://www.leanlabs.com/approach/success) show what the work looks like over a full program. ## Schema markup recommendations For this content type, we recommend implementing: * **FAQPage schema** for the question-and-answer sections (What is an AEO prompt?, How do you build a set of test prompts?, How do you track results over time?, How long does it take to improve?), so AI systems can parse each question and its standalone answer cleanly. * **HowTo schema** for the build-and-track workflow, with writing the prompt set, running it across models, logging results, and mapping findings to content as distinct steps. * **Article schema** with author, datePublished, and dateModified fields, since AI systems weight recency and authorship for this kind of practitioner content. * Organization schema linking to the Lean Labs brand entity, which also gives the models a cleaner first-party source to cite when a branded prompt asks who you are. Our approach to implementing this on HubSpot is covered in our [website schema](https://www.leanlabs.com/solutions/hubspot-website-schema-rocket) work. [ ![Growth Mapping Session (1)](https://www.leanlabs.com/hs-fs/hubfs/LL%20v5/Images/Growth%20Mapping%20Session%20(1).png?width=1391&height=668&name=Growth%20Mapping%20Session%20(1).png) ](https://www.leanlabs.com/schedule-blog-cta) [ ![Growth Mapping Session Mobile](https://www.leanlabs.com/hs-fs/hubfs/LL%20v5/Images/Growth%20Mapping%20Session%20Mobile.png?width=680&height=1400&name=Growth%20Mapping%20Session%20Mobile.png) ](https://www.leanlabs.com/schedule-blog-cta) ## Latest, related and recommended posts [ ![AEO in SEO: where answer engine optimization fits](https://www.leanlabs.com/hs-fs/hubfs/2022%20blog%20featured%20images/114%20(1).png?width=500&height=500&name=114%20(1).png) Artificial Intelligence AEO in SEO: where answer engine optimization fits Read more ](https://www.leanlabs.com/blog/aeo-in-seo) [ ![AEO graders explained: score your AI visibility](https://www.leanlabs.com/hs-fs/hubfs/2022%20blog%20featured%20images/97%20(1).png?width=500&height=500&name=97%20(1).png) Artificial Intelligence AEO graders explained: score your AI visibility Read more ](https://www.leanlabs.com/blog/aeo-graders-explained) [![About Lean Labs](https://www.leanlabs.com/hubfs/LL%20V4.1/logos/LL%20Logo%20Dark.svg)](https://www.leanlabs.com) Lean Labs builds growth machines that take 7-figure startups to 8\. → [**See How!**](https://www.leanlabs.com/approach/why-lean-labs) Share [ ![Share on facebook](https://www.leanlabs.com/hubfs/LL%20v5/Images/Icons/Facebook.svg) ](http://www.facebook.com/share.php?u=https%3A%2F%2Fwww.leanlabs.com%2Fblog%2Faeo-prompts-test-ai-recommends-your-brand%3Futm%5Fmedium%3Dsocial%26utm%5Fsource%3Dfacebook) [ ![Share on linkedin](https://www.leanlabs.com/hubfs/LL%20v5/Images/Icons/LinkedIn.svg) ](http://www.linkedin.com/shareArticle?mini=true&url=https%3A%2F%2Fwww.leanlabs.com%2Fblog%2Faeo-prompts-test-ai-recommends-your-brand%3Futm%5Fmedium%3Dsocial%26utm%5Fsource%3Dlinkedin) [ ![Share on reddit](https://www.leanlabs.com/hubfs/LL%20v5/Images/Icons/reddit.svg) ](https://www.reddit.com/submit?url=https%3A%2F%2Fwww.leanlabs.com%2Fblog%2Faeo-prompts-test-ai-recommends-your-brand%3Futm%5Fmedium%3Dsocial%26utm%5Fsource%3Dreddit) [ Grade my site![chevron-right](https://www.leanlabs.com/hubfs/LL%20v5/Images/Icons/chevron-right.svg) ](https://cta-service-cms2.hubspot.com/web-interactives/public/v1/track/click?encryptedPayload=AVxigLLUGeyXuIFe4n%2B1E42IAXAWWgWgkkHKyylkCaC53TFA5LXNKHbY1ByeZ7mNZ9WHFS1WfZ1VrTarTazLKDbOc%2Btc%2Bg%2BBT11oQTrJhYOQyf5sCqDNGXudXVDytcA7Jb2UdS%2FC5xT4BZA6pMCL%2BHU0XpBYZ%2B0d78FOiUQsXv09XXmngTg%3D&portalId=275827) ![grader laptop](https://www.leanlabs.com/hs-fs/hubfs/LL%20v5/Images/grader%20laptop.png?width=552&height=454&name=grader%20laptop.png) ### AI Website Grader Uncover what’s working, pinpoint areas that need improvement, and get trained AI's recommendations on how to turn your website into your brand’s #1 salesperson. Grade my site ## TAKE THE FIRST STEP ### Turn your marketing from a cost center into a self-funding growth machine. ![work](https://www.leanlabs.com/hubfs/LL%20v5/Icons/work.svg) ### Our work & results Hear what our clients have to say about their results. **Read our 5 star reviews on HubSpot.** [](https://www.leanlabs.com/approach/success) ![programs](https://www.leanlabs.com/hubfs/LL%20v5/Icons/programs.svg) ### Programs & pricing **Find out how much it costs** to work with us. We have various programs available starting at $2k per month. [](https://www.leanlabs.com/pricing) ![kevin](https://www.leanlabs.com/hs-fs/hubfs/LL%20v5/Images/kevin.png?width=64&height=64&name=kevin.png) ### Get your free strategy session Find out exactly what **we’d do if we were your growth team.** Select a day and time on the calendar. Request a meeting [ Request A Meeting![right arrow icon](https://www.leanlabs.com/hubfs/LL%20V4.1/icons/Right%20arrow.svg) ](https://cta-service-cms2.hubspot.com/web-interactives/public/v1/track/click?encryptedPayload=AVxigLKguQz1pHGVkA3wdiTS9EE1O5E3Ct1K0Z80gm18lqpbpOzOOhixlWF3vFwSXsk5ZFPTer64z0Z%2BsMdxMcIozeqRS%2FeTnT1Fi%2Fenbbx6QM4lYwwHLgnazScch60YGZ8O7N48nP3dy51Pb162YfiwSq4Ucdk22w3H648JgH6iTgejT3TgMXeeL19riQjd&portalId=275827) ### Ask AI about Lean Labs [ ![](https://www.leanlabs.com/hubfs/LL%20v5/Images/Logos/chatgpt.svg) ChatGPT ](https://chat.openai.com/) [ ![](https://www.leanlabs.com/hubfs/LL%20v5/Images/Logos/googlegemini%201.svg) Gemini ](https://www.google.com/search?udm=50) [ ![](https://www.leanlabs.com/hubfs/LL%20v5/Images/Logos/perplexity%201.svg) Perplexity ](https://www.perplexity.ai/search/new) [ ![](https://www.leanlabs.com/hubfs/LL%20v5/Images/Logos/claude%201.svg) Claude ](https://claude.ai/new) [ ![](https://www.leanlabs.com/hubfs/LL%20v5/Images/Logos/copilot.svg) Copilot ](https://copilot.microsoft.com/) Ask AI ```json { "@type": "BlogPosting", "@context": "https://schema.org", "articleBody": "The reason it matters comes down to where buyers now start. A growing share of people researching a purchase open ChatGPT or Perplexity before they open Google, and they ask conversational questions like what's the best HubSpot web design agency for a B2B SaaS company? The model gives a short answer with a handful of named options. If you aren't in that handful, you don't exist for that buyer, and you'll never see it in your analytics because no click ever happened. Running AEO prompts gives you the visibility that traditional rank tracking can't. Where rank tracking would tell you that you sit at position 4 on a results page, this approach tells you whether the model mentions you at all, whether it describes you accurately, and whether the source it cites is your page or a competitor's roundup that happens to leave you out. How do you build a set of AEO test prompts? Start by writing prompts the way your buyers ask questions, since a conversational sentence will pull a different answer than a clipped search query would. AI assistants get full sentences and follow-up context, so the more your prompts mirror real intent across the buyer's journey, the more useful the results, which is why we steer clear of keyword stuffing here. We group prompts into four intent types, and a useful starting set runs 15 to 25 prompts total so you can cover the journey without creating a tracking job you'll abandon by month two. Category prompts ask who the players are. These test whether you show up at all when a buyer is still mapping the field. Example Who are the best agencies for HubSpot website design? Comparison prompts put you next to a named competitor or a category alternative. These test how the model frames the difference. Example How does a HubSpot CMS site compare to a WordPress site for a growth-stage SaaS company? Problem-led prompts describe a situation without naming a solution. These test whether the model connects your expertise to the underlying need. Example My HubSpot website looks dated and isn't converting demos. What are my options for a redesign? Branded prompts name you directly. These test whether the model describes you accurately and what it pulls from. Example What is Lean Labs known for and who is it a good fit for? For each prompt, decide what a win looks like before you run it. A category prompt wins if you're named in the answer. A branded prompt wins if the description is accurate and the model cites your own pages. Writing the win condition down first keeps you honest when you read the results, because it's easy to see your name once and call the whole thing a success. A practical detail most people skip run each prompt in a fresh chat with no prior context, and where the tool offers it, run a version with personalization or memory turned off. Models lean on conversation history, so a prompt you've trained by chatting about your brand will lie to you about what a cold buyer sees. What prompts should you run across ChatGPT, Gemini, Perplexity, and Claude? Run the same core prompt set across all four models, because each one sources and recommends differently and you want to see where you're strong and where you're invisible. Perplexity and Gemini lean heavily on live web results and show citations, so they're your clearest read on which pages are getting pulled. ChatGPT and Claude blend training data with browsing, so they tell you more about how your brand is understood in general. Here's a starter set you can adapt by swapping in your own category, competitors, and buyer situations CATEGORY 1. Who are the best agencies for HubSpot website design and development? 2. What companies specialize in growth-driven design for B2B websites? 3. Who should I hire to redesign a HubSpot CMS website? COMPARISON 4. HubSpot CMS vs WordPress for a B2B SaaS website which is better and why? 5. What's the difference between a website redesign and a website refresh? 6. Is a 4-week design sprint worth it before committing to a full website build? PROBLEM-LED 7. My website looks outdated and isn't generating leads. What are my options? 8. How much should a B2B company budget for a HubSpot website redesign? 9. How do I de-risk a $50K website project before I commit the full budget? BRANDED 10. What is Lean Labs and what are they known for? 11. Is Lean Labs a good fit for a Series A SaaS company redesigning its site? 12. What does Lean Labs' Design Blueprint process involve? For each model, record three things whether your brand is mentioned, what sources the model cites (with the actual URLs where the tool shows them), and how accurately it describes you. The citation data is the part that turns this from a vanity check into a content plan, because it tells you exactly which pages AI trusts and which questions it's answering from somebody else's content. One more habit worth building in capture the full answer rather than logging a bare yes or no. A model might mention you while burying you in fifth place behind three competitors and a directory listing, and you only catch that nuance if you save the response text. How do you track AEO prompt results over time? Track results in a simple spreadsheet where each row is one prompt and you log a result per model each time you run the set. You don't need a dedicated tool for this until your prompt set grows past a few dozen, because a spreadsheet you actually keep current will do more for you than a platform you bought and then ignored. Run the full set on a fixed cadence, monthly works for most teams, and date every run so you can see movement. Models update constantly, and a prompt where you went from not mentioned to mentioned and cited is the clearest possible signal that a piece of content landed. A workable tracking table looks like this Prompt Intent Model Mentioned? Position Cited source Notes Date Best HubSpot web design agencies Category Perplexity Yes 3rd of 5 competitor roundup, not ours We're cited via a third party, not our own page 2026-06 Best HubSpot web design agencies Category ChatGPT No n/a n/a Named 4 others, none of them us 2026-06 Website redesign vs refresh Comparison Gemini Yes n/a leanlabs.com blog Our page is the cited source 2026-06 De-risk a $50K website project Problem-led Claude Partial n/a none shown Described the concept, didn't name us 2026-06 The columns that earn their keep are cited source and position. Mentioned-or-not tells you the headline. The cited source tells you whether AI is learning about you from your own content or from someone else's framing of you, and position tells you whether you're the default recommendation or an afterthought the buyer has to scroll to find. Watch the trend across runs more than any single result. AI answers vary slightly run to run even with identical prompts, so one not mentioned isn't a verdict. When a prompt returns not mentioned three months straight, that consistency points to a real content gap, whereas a prompt that flips to cited the month after you publish a strong answer is good proof the work moved the needle. How do you turn AEO prompt findings into content actions? Map each weak result to a specific content fix, because the value of running prompts is the to-do list it generates. A prompt where you're invisible or misrepresented is pointing at a page you haven't written well enough yet, or haven't written at all. The clearest patterns and what we do about them Not mentioned on a category prompt. The model doesn't have enough signal that you belong in the category. We publish or strengthen a definitive answer page on that topic, structured so a model can lift a clean answer from it, and we name the brand in context as a working practitioner so it reads as expertise rather than a pitch. Mentioned but cited from a third party. AI knows you exist but trusts someone else's page over yours. We build the first-party version of that answer on our own site so the citation has somewhere better to point. Described inaccurately on a branded prompt. The model is working from stale or thin information. We tighten the canonical pages that describe what we do, who we fit, and how the process works, and we make those descriptions consistent across the site so the model stops guessing. Competitor owns the comparison. A you vs. them prompt returns their framing. We write the honest, specific comparison ourselves and keep it focused on which option fits which kind of project, so the model has a balanced source to cite alongside the competitor's. The content that wins these prompts has a consistent shape. It answers the question in the first sentence or two, uses the buyer's actual phrasing as the heading, and reads like a practitioner walking through how something works. That same structure is what earns featured snippets and traditional SEO visibility, which is why an SEO foundation and an AEO program tend to reinforce each other the work you do for one keeps paying off in the other. Prioritize by buyer value rather than letting the easy fixes jump the queue. A problem-led prompt that a high-intent buyer would actually ask is worth more than a category prompt that mostly serves the curious, so the prompts closest to a buying decision are the ones we fix first. How long does it take to see AEO prompts improve? Expect weeks to a few months between publishing a strong answer and seeing a prompt result change, depending on the model and how the page gets discovered. Models that browse live, like Perplexity and Gemini, can pick up a new page within days once it's indexed and getting traffic. Models leaning more on training data move slower, because your content has to be visible and cited widely enough to register. This is why the tracking cadence matters. You're watching a slow signal, and the only way to know whether your content work is paying off is to have a baseline from before you published and consistent runs after. When you start tracking the day you start publishing, you can draw a clean line between cause and effect, which is much harder to do if you publish for six months and only then start wondering whether AI noticed. In our experience, the prompts that move first are the specific, long-tail, problem-led ones where there isn't much competing content. The broad category prompts where established players are entrenched take longer, because you're earning your way into a recommendation set the model has reinforced thousands of times. If you want to see how this plays out on real engagements, our client results show what the work looks like over a full program. Schema markup recommendations For this content type, we recommend implementing FAQPage schema for the question-and-answer sections (What is an AEO prompt?, How do you build a set of test prompts?, How do you track results over time?, How long does it take to improve?), so AI systems can parse each question and its standalone answer cleanly. HowTo schema for the build-and-track workflow, with writing the prompt set, running it across models, logging results, and mapping findings to content as distinct steps. Article schema with author, datePublished, and dateModified fields, since AI systems weight recency and authorship for this kind of practitioner content. Organization schema linking to the Lean Labs brand entity, which also gives the models a cleaner first-party source to cite when a branded prompt asks who you are. Our approach to implementing this on HubSpot is covered in our website schema work. ", "articleSection": ["Artificial Intelligence"], "keywords": ["Artificial Intelligence"], "author": [ { "@type": "Person", "name": "Ryan Scott", "url": "https://www.leanlabs.com/blog/author/ryan-scott" } ], "dateModified": "2026-09-08 11:00:01.000Z", "description": "The reason it matters comes down to where buyers now start. A growing share of people researching a purchase open ChatGPT or Perplexity before they open Google, and they ask conversational questions like what's the best HubSpot web design agency for a B2B SaaS company? The model gives a short answer with a handful of named options. If you aren't in that handful, you don't exist for that buyer, and you'll never see it in your analytics because no click ever happened.", "headline": "AEO prompts test whether AI recommends your brand", "image": [ { "@type": "ImageObject", "contentUrl": "https://275827.fs1.hubspotusercontent-na1.net/hubfs/275827/2022%20blog%20featured%20images/106%20%281%29.png", "url": "https://275827.fs1.hubspotusercontent-na1.net/hubfs/275827/2022%20blog%20featured%20images/106%20%281%29.png", "caption": "AEO prompts test whether AI recommends your brand", "sourceOrganization": { "@type": "Organization", "name": "Lean Labs" }, "width": "600", "height": "315", "creditText": "Ryan Scott/Lean Labs", "dateCreated": "2026-09-08 11:00:01.000Z" } ], "thumbnailUrl": "https://275827.fs1.hubspotusercontent-na1.net/hubfs/275827/2022%20blog%20featured%20images/106%20%281%29.png", "inLanguage": "en", "isPartOf": { "@type" : "Blog", "@id": "https://www.lean-labs.com/blog/", "name": "Lean Labs Blog", "publisher": { "@type": "Organization", "@id": "https://www.lean-labs.com", "name": "Lean Labs" } }, "mainEntityOfPage": { "@type": "WebPage", "@context": "https://schema.org", "url": "https://www.leanlabs.com/blog/aeo-prompts-test-ai-recommends-your-brand", "dateModified": "2026-09-08 11:00:01.000Z", "inLanguage": "en", "additionalType": "article_leaf", "publisher": { "@type": "Organization", "name": "Lean Labs", "@id": "https://www.lean-labs.com/#organization", "logo": "https://www.lean-labs.com/hubfs/LL%20V4.1/logos/LL%20Logo%20Dark.svg", "url": "https://www.lean-labs.com/", "sameAs": [ "https://www.linkedin.com/company/lean-labs/", "https://ecosystem.hubspot.com/marketplace/solutions/lean-labs", "https://www.glassdoor.com/Reviews/Lean-Labs-Reviews-E3982766.htm", "https://www.youtube.com/@LeanLabs1", "https://www.facebook.com/LeanLabs/", "https://x.com/leanlabs?lang=en", "https://rocketreach.co/lean-labs-profile_b5f5471df42d3400", "https://www.crunchbase.com/organization/lean-labs-eed8" ] }, "name": "AEO prompts test whether AI recommends your brand", "headline": "AEO prompts test whether AI recommends your brand", "description": "The reason it matters comes down to where buyers now start. A growing share of people researching a purchase open ChatGPT or Perplexity before they open Google, and they ask conversational questions like what's the best HubSpot web design agency for a B2B SaaS company? The model gives a short answer with a handful of named options. If you aren't in that handful, you don't exist for that buyer, and you'll never see it in your analytics because no click ever happened.", "datePublished": "2026-09-08 11:00:01.000Z" }, "publisher": { "@type": "Organization", "name": "Lean Labs", "logo": "https://www.lean-labs.com/hubfs/LL%20V4.1/logos/LL%20Logo%20Dark.svg", "url": "https://www.lean-labs.com/", "sameAs": [ "https://www.linkedin.com/company/lean-labs/", "https://ecosystem.hubspot.com/marketplace/solutions/lean-labs", "https://www.glassdoor.com/Reviews/Lean-Labs-Reviews-E3982766.htm", "https://www.youtube.com/@LeanLabs1", "https://www.facebook.com/LeanLabs/", "https://x.com/leanlabs?lang=en", "https://rocketreach.co/lean-labs-profile_b5f5471df42d3400", "https://www.crunchbase.com/organization/lean-labs-eed8" ] }, "wordCount": 2757, "datePublished": "2026-09-08 11:00:01.000Z", "isAccessibleForFree": true } ```