Here's the moment it clicked for me. I built the Superhuman Framework, and I started creating content around it. Then, the way I told it on HubHeroes:
"All of a sudden it started showing up in AI results when doing deep research... on their ChatGPT, their Perplexity, their places. I was like, wait, there's something to this. You can create something of value and it can be surfaced potentially quicker than it can be searched or found in search."
That's exactly why I chose the word "surfaced." As I said on that same episode, "Hence why I used the term surfaced when I actually did the 5S content framework." It happened again with a Tron Legacy article I wrote: "the thing was all of a sudden there because, again, it was in-depth. It was a pillar page piece of content and it was original."
So here's what I believe now. "I think it's about depth and originality."
This guide is the long version of that sentence. It's built on the method HubSpot Academy teaches in "Implementing AEO With HubSpot," stripped of the tool so it works with a spreadsheet and a browser, with outside research from Pew, Ahrefs, SparkToro, Google, OpenAI and Microsoft stacked on top.
What Is AEO, in One Answer?
Answer engine optimization (AEO) is the work of making your brand the one AI tools like ChatGPT, Gemini, Perplexity and Google's AI Overviews mention, cite and describe accurately when buyers ask questions. It builds on SEO, then adds answer-first pages, consistent facts across the web, third-party corroboration, and tracking mention rate instead of rank.
That's the definition. Here's why it matters: more of your buyers ask an AI assistant before they ever visit a website, and we want our answers to be the clear ones when that happens.
How to Use This Guide
You don't have to read this top to bottom. Pick your jump point:
- Beginner: AEO Foundations. How AI answers get built, why mentions beat clicks, a first-hour audit and five quick wins.
- Intermediate: Building the Practice. Readiness, prompt sets, page structure, off-site corroboration, technical basics, metrics and a 30-day plan.
- Advanced: Operating and Winning. Reading signals, triage, citations, experiments, sentiment, attribution and tools.
Part 1: AEO Foundations (Beginner)
New to this? Start here.
Why Are There So Many AEO Acronyms?
An "answer engine" is any tool that answers a question directly instead of handing you ten blue links: ChatGPT, Gemini, Perplexity, Microsoft Copilot, Claude, and Google's AI Overviews and AI Mode.
You'll also hear GEO (generative engine optimization), LLMO (large language model optimization) and more. On HubHeroes I put it this way: "you say AEO, another person I get on a meeting says GEO... Good god."
The label matters less than the job: getting your brand into the answer. In this guide, we'll just say AEO.
I even ran a HubHeroes episode trying not to say the acronym: SEO in 2026, and the acronym George refuses to say. Spoiler: I said it.
How Does an AI Answer Actually Get Built?
Here's what happens when a buyer types a question into an answer engine, in four moves.
1. The Engine Decides Whether to Search
OpenAI says "ChatGPT may search the web automatically when your question would benefit from current information."
Long, specific questions are where AI answers live. In Pew Research's study of real browsing (March 2025 data), searches of 10 or more words got a Google AI summary 53% of the time. Searches of one to two words got one just 8% of the time.
2. The Engine Fans the Question Out
The engine doesn't just run your buyer's exact words. It breaks the question into smaller ones.
OpenAI says ChatGPT search "typically rewrites your query into one or more targeted queries." Google's AI features page says AI Overviews and AI Mode "may use a 'query fan-out' technique, issuing multiple related searches across subtopics and data sources." The HubSpot Academy course puts it simply: "an answer engine breaks a single question into multiple sub-questions and searches for answers to each one."
So one buyer question becomes several searches behind the scenes, and every sub-question is a chance for your page to get pulled in.
3. The Engine Pulls Passages, Not Pages
The engine doesn't read your page top to bottom. It grabs pieces. Perplexity's engineering team describes breaking documents "into self-contained spans, each of which can be individually retrieved and ranked." Microsoft says AI assistants "break content into smaller, usable pieces."
The Academy course says engines want content they can "extract cleanly, quote accurately, and attribute confidently." A brilliant answer buried in a rambling section is an answer nobody retrieves.
4. The Engine Looks for Consensus
This move changes your strategy the most. HubSpot's AEO product lead says answer engines "try to reach consensus by actually triangulating a bunch of different data sources."
In plain words, the engine checks you against everybody else: review sites, Reddit threads, YouTube videos, podcasts and media coverage. The answer reflects the pattern across all of them.
And the engines don't lean on the same sources. In Profound's 680 million citation dataset (2025), Wikipedia led ChatGPT's top sources and Reddit led Perplexity's. Each engine is its own room.
Why Do Mentions Matter More Than Clicks Now?
The clicks are shrinking. In Pew's March 2025 browsing data, humans clicked a regular search result on 8% of visits when Google showed an AI summary, versus 15% without one. They clicked a link inside the summary just 1% of the time. Ahrefs found the top organic result's click-through rate was 58% lower on keywords with an AI Overview (December 2025 data, 300,000 keywords).
I see it in my own numbers. My HubSpot Hubs complete guide pulled about 105,000 impressions in a quarter and 27 clicks. That's Search Console, August 2026. My read when I shared it on HubHeroes: "something is or somebody is" asking, and "the historical homepage in the first five results is going away."
So what's the prize now? HubSpot's Aja Frost, who led the team that made HubSpot visible in answer engines, said it straight: "this isn't about getting traffic anymore... what really matters is getting mentioned." The mention drives the purchase, she said, "probably not within that same session, but ultimately."
One hopeful signal: Seer Interactive found that when an AI Overview showed and the brand was cited, organic click-through was 35% higher than when it wasn't (Q3 2025, 42 client organizations). Seer is careful to say causation isn't clear, so treat it as a signal worth watching, not proof.
The hard truth: on unbranded "best X for Y" questions, only about 4% of AI citations point to the brand's own website, according to Profound's study of 27 million citations (January 2026). The rest are media, review sites, communities and other brands.
I said it back in 2025: "content is the fuel for AI, large language models, they're going to quote, they're going to cite, they're going to learn from these authoritative pages, these authoritative websites, hopefully your website. And at the end of the day, if your brand is not publishing, it's invisible to any of those possible answers."
“"I believe we're going into a world where it's less about traffic and more about impact with the content."”
Is SEO Dead?
No. And I'll say it the way I said it on HubHeroes: "You should not be having a conversation of content is dead or that SEO is dead. SEO is part of what's happening."
The full story is in the demise of SEO and the morning I almost let everything burn.
Two reasons SEO still matters for AEO:
- Ranking still feeds Google's AI answers. Ahrefs found only about 12% of URLs cited by ChatGPT, Gemini and Copilot rank in Google's top 10 for that prompt (August 2025), but Google's AI Overviews tracked much closer to the organic results.
- Your ranking pages are your best raw material. The Academy course says "pages that rank well in organic search are more likely to be cited by answer engines."
The HubSpot Academy course nails it: "AEO isn't a replacement for strong content strategy. It's a higher bar for it."
I said it this way on the show: "I hope people are understanding how directly tied AEO is to SEO and how human first, value first, content driven this whole conversation needs to be."
Why Is They Ask, You Answer the Best AEO Strategy?
Because answer engines answer questions, and TAYA was always about answering your buyers' real questions better than anybody else.
At INBOUND, I watched Yamini on stage, and I started laughing. I leaned over to the human next to me and said, "They Ask You Answer is the best AEO." Then, on HubHeroes, I said the full version:
“"They Ask You Answer is the best AEO strategy that we never knew was gonna be an AEO strategy."”
TAYA asks you to publish on cost, problems, comparisons, best-of lists and reviews. Now look at what buyers type into ChatGPT: "How much does it cost?" "What's best for a team like mine?" "X vs Y?" Same list. Here's what a TAYA cost article looks like in practice: How much do HubSpot Hubs cost?
Back in 2025 I said, "I'm not trying to be a homer of They Ask You Answer." But "it must be mapped to the buyer's pain points and their business goals." That part hasn't changed.
Go further with how They Ask You Answer became the blueprint for AEO and our breakdown of the They Ask You Answer approach to inbound, with examples.
What Should You Do in Your First Hour?
Run a manual audit: a spreadsheet, three browser tabs and an honest hour. It's the audit the HubSpot Academy course teaches, and it works for a team of one.
Step 1: Write 10 Real Buyer Prompts
Write them the way a buyer would ask a trusted advisor, spread across the journey:
- Awareness: "How do I solve [problem]?"
- Consideration: "What are the best [solutions] for [type of business] with [situation]?"
- Decision: "[Your brand] vs [competitor], which is better for [use case]?"
The Academy course says to pay the most attention to decision-stage prompts.
Step 2: Run Them in Three Engines
Run every prompt in ChatGPT, Perplexity, and Google with AI Overviews or AI Mode.
Step 3: Log What You See
For each answer, write down:
- Which brands get named.
- Whether you appear at all.
- How you're described: top pick, one of several, or absent.
- Which sources get cited.
- Anything flat-out wrong about you (old pricing, a discontinued service, the wrong city).
Step 4: Expect It to Wobble
Run the same prompt twice and you'll get a different answer. The Academy course warns that "AI responses vary by session, by user context, and even by time of day," so read your audit as "general patterns, not definitive rankings."
The research backs that up. SparkToro and Gumshoe ran the same prompts 2,961 times (January 2026). There was less than a 1 in 100 chance of getting the same list of brands twice, and about a 1 in 1,000 chance of the same order. But the set of brands that keep showing up was much more stable. Watch the set, not the order.
What Are the Five Quick Wins?
Here are five fixes you can start this week.
1. Put the Answer First
Open every important page with a direct answer to its main question. HubSpot's AEO product lead called this the single most important page change: "a very direct answer to the question... at the top of the page."
If the answer is buried, the Academy course warns, the engine "may pull from a competitor's page that delivers the answer faster."
2. Add Visible FAQs That Answer Real Buyer Questions
Not filler. Real questions from sales calls and support tickets, each with its own heading and a short, honest answer. Every FAQ is a sub-question the engine might go looking for.
3. Don't Block AI Search Bots
This is the strongest evidence in the whole guide. OpenAI's bots documentation says: "Sites that are opted out of OAI-SearchBot will not be shown in ChatGPT search answers."
Every major AI company now separates training bots from search bots. Blocking a training bot (GPTBot, ClaudeBot, Google-Extended, Applebot-Extended) doesn't remove you from their AI search answers. Blocking a search bot does. And Google-Extended doesn't control AI Overviews or AI Mode; those run on Googlebot. The full bots table is in Part 2.
4. Fix Inconsistent Facts Across the Web
If your pricing says one thing on your site and another on a directory, the engine has to triangulate a mess. Pick the true version and update it everywhere you control: your site, Google Business Profile, directories, partner pages and social bios.
5. Refresh Pages That Already Rank
I called this one out on HubHeroes: "maybe it's a nice little schema and some FAQs that needed added to an article that was already ranking fine for search. Now all a sudden, boom, it's getting surfaced in AI. So like we've talked about historical optimization for SEO, but how about now historical optimization for AI?"
Start with the pages already pulling impressions. The Academy course agrees: if you already have a page on the topic, "revision is almost always the faster path."
Part 2: Building the Practice (Intermediate)
How Ready Are You for AEO? The Three-Part Assessment
The HubSpot Academy course uses a readiness assessment with three dimensions: content maturity, competitive landscape, and buyer query clarity. It works, in the course's words, for "a two-person marketing team or a 50-person content organization."
Content Maturity, Layer 1: Your Organic Foundation
Pull your top 10 to 15 pages by organic traffic or impressions from Search Console. If very few pages rank, the course is blunt: "build organic presence first."
Content Maturity, Layer 2: Extractability
Now read those pages like a machine would, using the checklist below. A page that ranks well but fails the structure check is, in the course's words, "strong raw material that needs reformatting."
Competitive Landscape
Your AEO competitors are whoever shows up when buyers ask. The course says that "often also includes media publications, industry analysts, review platforms, and educational resources."
Pick 3 to 5 competitors your buyers actually compare you against, then look at the sources cited next to them.
Buyer Query Clarity
The course calls this "the one most marketers underestimate." Keyword research tells you volume. Buyer query research tells you how humans phrase problems, "the way they would talk to a trusted advisor."
Where to find those questions:
- Support tickets, chat transcripts and CRM notes.
- 3 to 5 recorded sales calls, with the questions written down word for word. The course calls those "gold."
- Reddit, Quora and industry communities.
Aim for 10 to 15 high-intent questions sorted by journey stage. A question that shows up in more than one place is a strong signal.
Reading the Three Together
- You rank but don't get cited? That's a formatting problem.
- You've got no content for real buyer questions? That's a coverage gap.
- Your content is strong and well-structured, and you still lose? That's authority, original data, or a missing tie back to your brand.
Small teams usually find gaps in all three; start narrow, with high-intent, low-competition questions. Enterprise teams usually have the raw materials, but "the connective tissue between them is missing."
How Do You Build a Prompt Set That Tells the Truth?
Your prompt set is the list of questions you'll track over time. As my grandpa used to say, garbage in, garbage out.
Start From Real Questions, Not AI-Generated Ones
HubSpot's AEO product lead suggests building a matrix of your products against your buyer types (title, location) "to create coverage," and warns against accepting AI-generated prompts at face value.
Use the Frost and Hope Method for Size
HubSpot's Aja Frost described the method her team used with Rory Hope: about 20 questions per main offer, five per buyer-journey stage, shaped by your persona. Her example starts with a sleepless buyer whose neighbors "have been doing construction for two years," and ends with one white noise machine versus another.
Start with consideration-stage prompts, where the engine is actively choosing who to recommend.
Run Every Prompt Through Three Checks
The Academy course gives three questions. Any "no" means rewrite it or cut it:
- Would a real buyer type this?
- Does it reflect a buying decision?
- Is there a competitive angle?
Good Prompts vs Bad Prompts
| Bad prompt | Why it fails | Better prompt |
|---|---|---|
| What's marketing automation? | No brand gets recommended, so visibility reads zero | Best marketing automation platform for a small B2B SaaS team |
| Best CRM | Too broad, no buyer, no situation | Best CRM for a small home services company that books jobs by phone |
| Is [your brand] good? | No competitive angle | [Your brand] vs [competitor] for a mid-size manufacturer |
| Top marketing agency | Nobody talks like this | Which agency can help a small team fix a messy HubSpot portal |
The first row comes from the Academy course. If no brands show up, make the prompt more specific. And treat the list as "a living asset."
How Should You Structure Pages for Passage Retrieval?
Engines pull passages, not pages, so every section has to earn its spot on its own. Here's the checklist, built from the Academy course and the outside research.
| Checklist item | What it means | Why it matters |
|---|---|---|
| Answer first | The first sentence or paragraph answers the main question | Engines favor the page that answers fastest |
| Self-contained sections | Each section makes sense alone, no "as we said above" | Passages are judged one at a time |
| Explicit structure | Descriptive headers, lists and tables, not "Our approach" | Clear structure is easier to extract and quote |
| Original data | Your own stats, results, CRM trends or survey numbers | Engines look for net new information to cite |
| FAQ sub-questions | Real questions with their own headings and short answers | Matches query fan-out sub-questions |
| Tie back to your brand | Connect the answer to what you actually offer | Otherwise you teach the engine and it recommends someone else |
| Visible updated date | Show both a published and a last-updated date | Freshness is one of the two big reasons pages get displaced |
| Light and fast | Fast load, real text, nothing important hidden in tabs or images | Hidden content can be missed |
Front-load the page. Kevin Indig's analysis of ChatGPT citations (February 2026) found 44.2% came from the first 30% of the page, and 24.7% from the final third. In the same study, cited text was nearly twice as likely to use definitional phrasing like "X is..."
Add proof. The 2024 academic GEO study from Princeton and IIT Delhi found that adding quotations, statistics and cited sources raised visibility by up to about 40% in lab tests. Keyword stuffing did little. Those were 2023-era engines, so treat it as direction, not a promise.
Don't chop the page into bits. Google's Danny Sullivan, on chopping content into "bite-sized chunks" for LLMs (January 2026): "we don't want you to do that." Write clear, well-headed sections for humans.
Microsoft recommends clear headings, Q&A formats, lists and tables, and warns that content hidden in tabs, PDFs or images may be missed. HubSpot's AEO product lead adds: keep the page "really light and fast... The crawler gets bored."
Why Is Original Data Your Moat?
The Academy course says "answer engines are looking for net new information to cite." It doesn't need to be a big survey. CRM insights, customer success trends and campaign results all count. So do case studies that carry real numbers.
This is where depth and originality stop being a slogan. Back in 2025 I called out the shift: "AI answer engines and zero click have absolutely shrunk the top of the funnel... The mindset of blog every day, this idea of quantity over quality, I think has gone away."
“"Five pieces of slop a day is worse than one real piece a week."”
That's my line on volume. My number is 3 to 5 original pieces a week, per company. Consistency beats volume. If you want depth that compounds, organize it as a topic cluster so every page supports the others, and spend the time you save on creating less and promoting more.
How Do You Get Corroborated Off Your Own Site?
The engine triangulates, so the work moves beyond your website.
Mentions Are the New Backlinks (Mostly)
Ahrefs studied 75,000 brands (2025) and found AI Overview visibility tracked branded web mentions (0.664) about three times more closely than backlinks (0.218). That's correlation, not proof, but it points the same way: be talked about where engines read. A later Ahrefs analysis found YouTube mentions tracked even more closely.
Where Corroboration Comes From
- Reviews on the platforms your buyers use.
- YouTube videos that answer buyer questions, with transcripts.
- Podcasts, where guest spots put your ideas in places engines index.
- Media and roundups that already get cited for your topic.
- Communities like Reddit, forums and LinkedIn conversations.
Communities: Participate as Yourself
A guardrail: Reddit is cracking down on brands posting "stealth marketing" aimed at AI chatbots, according to Bloomberg (July 2026). And Google's spam policy now covers "attempting to manipulate generative AI responses in Google Search."
So show up as yourself, disclose who you work for, and help even when you're not the answer. Never use fake accounts.
How Do You Make Your Brand Easy for Engines to Understand?
Consensus is hard when your own facts disagree. You want the engine to know exactly who you are and find the same facts everywhere it looks.
- Build a real About page. Who you are, what you do, who you serve, who leads the company. Plain text.
- Add Organization schema with sameAs. Google's Organization structured data lets you list your official name, logo and the profiles that belong to you (LinkedIn, YouTube, directory listings) with
sameAslinks. - Use the same facts everywhere. If a human could find two versions, the engine will too.
HubSpot's Aja Frost shared a story. ChatGPT was giving out incorrect HubSpot pricing. Her team published "a series of pricing blog posts... incredibly literal... like a catalog," and "ChatGPT visited that series within hours of us publishing it, and the accuracy of pricing information went up immediately." Her lesson: "If there's anything that's incorrect or missing from the narrative, put it on your blog." That's one vendor's anecdote, but the principle travels.
What Are the Technical Basics?
The AI Bots Table
"Search" bots decide whether you show up in answers. "Training" bots are a separate business choice. "User-triggered" fetchers visit a page when a human asks a question.
| Bot | Company | Job | What to do if you want citations |
|---|---|---|---|
| OAI-SearchBot | OpenAI | Search: ChatGPT search answers | Allow, and allow its IPs at your CDN |
| GPTBot | OpenAI | Training | Business choice, separate from search |
| ChatGPT-User | OpenAI | User-triggered fetch | Allow at the firewall level |
| PerplexityBot | Perplexity | Search: Perplexity results | Allow |
| Perplexity-User | Perplexity | User-triggered fetch | Allow at the firewall level |
| Claude-SearchBot | Anthropic | Search quality for Claude | Allow |
| ClaudeBot | Anthropic | Training | Business choice |
| Claude-User | Anthropic | User-directed fetch | Allow |
| Googlebot | Search, including AI Overviews and AI Mode | Allow | |
| Google-Extended | Gemini training and grounding token | Business choice, doesn't affect AI Overviews or AI Mode | |
| Bingbot | Microsoft | Bing index and Copilot | Allow |
Check Your CDN, Not Just robots.txt
Your robots.txt can say "welcome" while your CDN or firewall slams the door.
Cloudflare changed its AI bot defaults several times (July 2025, September 2026). Since September 15, 2026, new ad-supported domains block "Agent" traffic on ad pages by default. OpenAI also requires that your host or CDN allow its search bot's IP addresses, because a firewall block overrides a friendly robots.txt. Log into your CDN and look.
Bing Webmaster Tools and IndexNow
"ChatGPT runs on Bing" is out of date as a sole strategy. 2026 research observed OpenAI's own index doing most of ChatGPT's retrieval. Still, verify your site in Bing Webmaster Tools and turn on IndexNow, because Copilot runs on Bing.
Schema: My Honest Update
In 2025 I said, "If you're not paying attention historically to schema, schema matters." Here's where the evidence has taken me since.
Google says "There's also no special schema.org structured data that you need to add" for AI Overviews or AI Mode, and no "AI text files." Microsoft recommends JSON-LD schema to help its AI understand content. And in a 2026 Ahrefs before-and-after test on 1,885 pages, adding schema didn't move AI citations in any meaningful way. One caveat: those pages were already cited.
Updated take: schema helps engines understand who you are, especially Organization and sameAs, and it costs little. It won't get you cited by itself. Visible text beats hidden markup.
llms.txt: Ship It for Agents, Then Stop Thinking About It
Ahrefs (June 2026) found that of the sites publishing one, 97% got zero requests for it in a month. Google says llms.txt "will neither harm nor help" in Search.
So llms.txt is for AI agents, not citations. That fits how I think about agent-first websites: "Start with what agents can read before what they can do." Ship one, then move on.
Which AEO Metrics Actually Matter?
Five metrics, with definitions from the HubSpot Academy course. You can track all of them in a spreadsheet.
- Brand visibility. The percentage of answers that mention you, across your tracked prompts.
- Sentiment. Whether those mentions are positive, neutral or negative.
- Share of voice. Your mentions divided by all mentions of your brand plus your tracked competitors.
- Citations. Which sources the engine links to when it answers, and whether any are yours.
- AI referrals. Visits that arrive from an AI tool, measured in your analytics.
HubSpot's AEO product lead framed the first three this way: "We want the answer engines to be our best sales rep." Does the rep mention you, what does it say, and how often compared to everybody else?
Mention Rate, Not Rank
Don't report an "AI rank." With less than a 1 in 100 chance of the same brand list twice, a rank from one run is noise dressed up as a metric.
Report mention rate across many runs. Your baseline, per the Academy course: the percentage of tracked prompts where you appear.
What Goals Should You Set?
The Academy course's first rule: don't borrow someone else's goals.
Dense Markets vs Sparse Markets
- Dense market: lots of brands cited (think "best CRM for small business"). Share of voice is your core metric, and the aim is to "appear more consistently in more favorable contexts."
- Sparse market: thin coverage, like estate planning attorneys in one city, where the answer might name one or two firms, or none. Your job is "to be accurate, present, and characterized positively."
Enterprise vs Small Business
Enterprise teams fight for category authority. Small businesses fight for visibility first, often on local questions. In the course's words, "The mistake is borrowing the other's playbook."
Write Goals You Can Actually Evaluate
Four rules from the course:
- Watch trends, not absolute scores.
- Connect to outcomes leadership already tracks, like pipeline and brand awareness.
- Make it specific enough to evaluate in 30 to 90 days. "'Show up more in AI answers' isn't a goal. 'Increased citation frequency for three target queries by the end of the next quarter' is."
- Expect to iterate.
What Does a 30-Day AEO Plan Look Like?
AEO is a long game. More like a crock pot, less like a microwave. Here's a 30-day plan built on the Academy course's sequence.
Before day one. Finalize 3 to 5 competitors, document buyer questions by stage, and brief your stakeholders.
Week 1: Baseline and refine. Run your prompt set and record the baseline. Resist acting on early data. Fix prompts that return no brands or don't reflect a real buying decision.
Week 2: Restructure your strongest pages. Start with the pages that already rank.
Week 3: Fix facts and access. Clean up inconsistent facts across the web. Check your robots.txt and CDN settings. Verify Bing Webmaster Tools and turn on IndexNow.
Week 4: Track and compare. Keep running the same prompts and compare direction, not single days.
The course offers three success indicators for the month:
- Restructure 5 to 10 high-traffic pages in 30 days (for a small team, 2 to 3 is the realistic version: "depth over breadth").
- Run 4 consecutive weeks of tracking.
- Appear for at least 2 buyer queries where you're absent today.
And one big warning: don't set a visibility-score target for the first 30 days. The course says the metrics "fluctuate meaningfully," so you can't tell your work from platform shifts or noise.
The line to remember from this part of the course: "The most important thing about a 30-day plan is that it's designed to be revised at day 31."
Part 3: Operating and Winning (Advanced)
Now the job is reading signals, picking battles and running clean experiments.
How Do You Read the Metrics Together?
The Academy course's first warning: don't treat these numbers like domain authority. If visibility "dropped two points on a Tuesday, that's probably not" meaningful.
Visibility and sentiment "can move in completely opposite directions." You can be cited constantly and described as "expensive or difficult to work with." So read them as a pair.
| Positive sentiment | Negative sentiment | |
|---|---|---|
| High visibility | Protect it: stay consistent and don't over-optimize | Urgent: investigate the source material before chasing more coverage |
| Low visibility | Expand coverage: get into more prompts | Hardest: fix the characterization first, because more coverage will only amplify the problem |
Share of voice is relative, so a drop "could just mean a competitor started appearing more."
Watch competitors, too. One with much higher visibility shows you what engines reward. One that never appears means low visibility or the wrong competitor set.
Sentiment is a leading indicator: visibility can hold steady while sentiment "quietly erodes." Prompt-level negativity tells you exactly which story to fix.
Protect, Improve or Pursue: How Do You Triage Prompts?
Every tracked prompt goes into one of three buckets, straight from the Academy course.
- Protect. You show up consistently, favorably, across engines. Keep those pages current and well-linked, but "don't artificially inflate your content freshness."
- Improve. You show up but underperform. Compare the content cited for you against what's cited for competitors, then optimize.
- Pursue. You're absent, but the question is real. "Not every gap is worth closing." Pick 2 to 3 high-value prompts where owned content can realistically earn a place, start there, and reassess in about a month.
Prompts move between buckets, so re-triage regularly.
How Do You Analyze Citations?
"The brand whose content gets cited is the one whose perspective reaches the buyer," the course says.
Two Rates, Read Together
| Owned citation rate | Brand mention rate in cited sources | What it means |
|---|---|---|
| High | High | Strongest position: your content is source material and third parties mention you |
| Low | High | You're in the conversation, but someone else controls the story |
| Low | Low | You're largely absent from AI answers in your category |
Fix Format Mismatch
If comparison articles and listicles dominate citations for your prompts and all you publish is blog posts, you've got a format mismatch. Format preferences also vary by engine, so check per engine.
On listicles: Lily Ray found Google's AI Overviews often cited brands' self-promotional "best of" lists but left the publishing brand out of the recommendation 69% of the time. So if you build comparison pages, make them honest: fair criteria, stated weaknesses, and clear authorship. No "we're number one" lists. Our HubSpot alternatives guide is one example of the honest kind.
Get Into the Sources Engines Already Cite
The course calls this "one of the most effective ways to influence how AI describes you": find the third-party sources that keep getting cited for your prompts and earn a mention there. Prioritize by visibility gap, not citation gap. Start from your flagship offer, ask which sites get cited for it, and go after those lists, roundups and publications.
HubSpot's Aja Frost said "Blog posts are overrepresented as a citation type" in HubSpot's data, and added: "Don't spin down your blog. Your blog is actually like your megaphone to LLMs." That's HubSpot's own data, so weigh it accordingly. Her shorthand: "B to B to C, business to bot to consumer."
Why Do You Get Displaced?
You were cited last month, and now a competitor took your spot. The Academy course says citation displacement mostly comes down to two things: freshness and structure.
Freshness. No visible update date, or stats from years ago, "signals... the content may not reflect current thinking." Show both a published and a last-updated date. Refresh stats and tool references.
Ahrefs (17 million citations, July 2025) found AI assistants cite content that's on average 25.7% fresher than what ranks in regular search. But the average cited page is still about 2.9 years old. Freshness is a tilt, not a rule. Update the substance, not just the date.
Structure. The course's test: "whether someone asking that exact prompt would find a direct answer within the first two paragraphs."
The best calibration: open the winning page next to yours, read both openings, and note what theirs does that yours doesn't.
Also ask whose content is winning: "A competitor's owned content winning a prompt is a different problem than a third-party publisher dominating it." You can potentially outwrite the first. The second takes outreach and authority.
How Do You Run AEO Experiments?
Updating a page and running an experiment are different things. Change five things at once and, in the course's words, "there's no way to know which change drove the result."
The One-Variable Method
- Record the baseline and the date. Which prompts, which engines, current mention rate.
- Make one change. A restructured opening, a new section for an adjacent question, or an updated stat.
- Leave it alone. No tinkering for a few weeks.
- Compare. Same prompts, same engines, same number of runs.
"No movement... suggests the variable you changed wasn't the issue. This is important information too."
How Many Runs Are Enough?
Graphite found 10 runs per prompt is enough for a quick estimate, with about a 5.6% average error on visibility (April 2026). Don't judge anything on one run.
What's Normal Movement?
HubSpot's AEO product lead says to expect about 5 to 10% normal movement week to week. Wait for one to two weeks of stability before acting. A 20 to 30% drop is different: check for bot blocking, an indexing problem, or a big engine change.
A useful control: track a few category and competitor prompts you're not working on. If they all moved too, it probably wasn't you.
Treat AEO Advice as a Hypothesis
The course is honest here: "Some of what circulates as established guidance right now will turn out to be incomplete, overstated, or only true in specific contexts." That includes this guide. Your own data beats outside commentary.
How Do You Fix Negative Sentiment at the Root?
HubSpot's Aja Frost was blunt: "If you're mentioned 10 times, but every time an answer engine is saying, don't use this product... you would rather have zero mentions."
The course notes that negative sentiment "tends to concentrate on the areas of the product or service that real customers genuinely have issues with." Frost said the same objections showed up on HubSpot's sales calls, because "the answer engines go and train where the people are talking."
So here's the mistake: thinking you can publish more content to drown it out. You can't. In the course's words, "You have to actually address the root cause," then tell the world.
Frost's example was Domino's: "their pizza sauce used to taste like ketchup, it doesn't anymore." They fixed the product, then they told everybody.
Compare that with the pricing posts: when facts are missing, publish them. When the experience is bad, fix the experience first, then publish what changed.
How Do You Measure AEO and Attribute It?
Frost said attribution misses "the entire middle of their journey." So build a stack of signals and read them together.
GA4: The AI Assistant Channel
GA4 added a default "AI Assistant" channel on May 13, 2026. It isn't retroactive, and it misses visits with no referrer, so treat it as a floor.
For your own custom channel, use Session source, "matches regex," placed above Referral. Here's a regex published by Terminus (July 2026); check it against your own referral list:
^(claude\.ai|perplexity\.ai|www\.perplexity\.ai|meta\.ai|you\.com|poe\.com|chatgpt\.com|chat\.openai\.com|gemini\.google\.com|bard\.google\.com|copilot\.microsoft\.com|chat\.deepseek\.com|grok\.com)$
ChatGPT also adds utm_source=chatgpt.com to the links it sends.
Google Search Console: Generative AI Report
Search Console's generative AI performance report shows impressions in AI Overviews and AI Mode. It launched in June 2026 and rolled out to all sites worldwide as of August 31, 2026. It's impressions only: no clicks, no queries.
To find question-shaped searches in your regular performance report, filter queries with this regex for 8 or more words:
^(\S+\s+){7,}\S+$
That finds prompt-like questions. It isn't proof of AI Mode traffic.
Bing Webmaster Tools: AI Performance
Bing's AI Performance report (February 2026) shows citations across Copilot, including the grounding queries: the only first-party report that shows the queries behind AI citations.
The Floor, Not the Ceiling
The Academy course says to think of AI referral traffic as "the floor, the most conservative possible count of AI's influence." Read AI referrals next to branded search, direct traffic, brand visibility and citations. When they move together, they tell "a much more compelling story than any single number in isolation."
Ask Humans How They Found You
Frost said the most successful attribution technique she's seen is self-reported attribution. Add "How did you hear about us?" to your forms and your sales calls. Make it an open text field.
This ties right back to what I said about the 5S framework: "while your traffic might be down, look at your referrals." And it's why I push back on humans "clinging to vanity metrics" without figuring out how content "is influencing the actual pipeline."
Does AI Traffic Convert Better?
It's contested. Adobe's US retail data (July 2026) found AI-referred visits converted 60% better, after converting worse in early 2025. An academic study of 973 stores (2024 to 2025) found ChatGPT traffic converted below organic search. Measure your own.
Which AEO Tools Should You Use?
Start free. Add a tracker if the free reports leave you guessing.
Free, First-Party Tools
- Google Search Console generative AI report: impressions in AI Overviews and AI Mode.
- Bing Webmaster Tools AI Performance: Copilot citations plus grounding queries.
- GA4 AI Assistant channel: sessions from AI tools.
- A spreadsheet and three browser tabs: your manual prompt runs.
Paid Trackers
These run your prompts on a schedule and track visibility, sentiment, share of voice and citations. We're not ranking them, and prices change, so verify before you buy.
| Tool | Notes (as of October 2026) |
|---|---|
| HubSpot AEO | $50 a month standalone, also in Marketing Hub Pro and Enterprise; tracks ChatGPT, Gemini and Perplexity; free AEO Grader |
| Peec AI | From $95 a month |
| Otterly.AI | From $29 a month |
| Scrunch | From $250 a month, billed annually |
| Profound | Enterprise pricing, with a trial |
| Semrush AI Visibility Toolkit and Ahrefs Brand Radar | Available; check current pricing with the vendor |
If you're already on HubSpot, read our breakdown of HubSpot's AEO tool in Marketing Hub. The tool runs the prompts. You still have to read the signals.
What Happens When Owned Content Plateaus?
At some point your pages are answer-first, fresh and structured, and the numbers stop climbing. The Academy course: "Owned content has a ceiling."
If several experiments haven't moved your citation share, "the issue probably isn't your content. It's your authority... When your owned citation rate has plateaued despite consistent effort, outreach, PR, and community presence stop being optional."
The course lists three action types, and this is when the second and third take over:
- Owned: create or update pages.
- Social amplification: some engines cite social, community and forum content, so this matters more when your citation data shows social sources.
- Outreach: pitch journalists, contribute to roundups and build publisher relationships. It's slower, but it "can produce durable citation gains." No PR budget? Start with how to become the go-to expert in your niche.
What's Next: Ads and Agents
Dated October 2026, because this moves fast.
Ads in answers. OpenAI began testing ads in ChatGPT on February 9, 2026, and has expanded them to more countries since. OpenAI says, "Ads do not influence the answers ChatGPT gives you." Organic answers are still the job.
Agents. ChatGPT agent (July 2025) browses and takes actions. Google's AI Mode runs agentic tasks like restaurant booking. The practical move: make your prices, policies and booking flows readable as plain text.
I won't repeat my agent-first websites piece here. The short version: "Human-first content, agent-first infrastructure." If you want help making your site readable to agents, that's our agent readiness work.
How Do You Run a 20-Minute Weekly AEO Review?
The Academy course's 20-minute weekly review answers three questions: Is visibility trending the right way? Have competitive positions shifted? Is there an action worth taking this week?
- Check visibility and sentiment for the last 7 days, then compare against the last 30 to separate trend from noise.
- Check share of voice. "A lower share of voice doesn't automatically mean your AEO strategy isn't working."
- Check citations. Any new domain in the top list? A competitor domain climbing? A shift between channels?
- Check competitor trajectory over 30 days. A competitor who's always outranked you is context. One whose visibility has been climbing steadily is a different kind of signal.
- Revisit the prompts you acted on. Did your experiment move anything yet?
- Read by journey stage. For awareness and consideration prompts, sort by low visibility. For evaluation and decision prompts, which usually name your brand, read how you're described.
- Leave with one or two actions, each with an owner.
"The value of this review isn't any single session, but the consistency of doing it over time."
Every quarter, zoom out. Does the prompt set still match what buyers ask? Is owned content still the right lever? "Every slot should earn its place."
AEO Myths and Contested Claims
What the evidence says, as of October 2026:
| Claim you'll hear | What the evidence says |
|---|---|
| An llms.txt file gets you cited | Ahrefs found 97% of sites with one got zero requests for it in a month. Ship it for agents, not citations |
| Schema gets you cited | Google says no special schema is needed, and a 2026 Ahrefs test found schema didn't move citations. It helps engines understand who you are |
| Track your AI rank | SparkToro found less than a 1 in 100 chance of the same brand list twice. Track mention rate across many runs |
| Chunk everything into tiny pages | Google's Danny Sullivan said "we don't want you to do that." Write clear, self-contained sections on solid pages |
| Publish 10x more content | Depth and originality beat volume. Original data and real answers get cited |
| Self-promotional best-of lists work | Lily Ray found AI Overviews often cited them but left the publishing brand out of the recommendation 69% of the time |
| Block all AI bots to protect your content | Blocking search bots removes you from those answers. Blocking training bots is a separate choice that doesn't |
| SEO is dead | Ahrefs found Google's AI Overviews track much closer to organic results than other engines do. AEO is a higher bar on SEO |
Should You Gate Content for AEO?
My position is ungated by default: "An article is never worth an email address." Gate only when there's a real exchange of value, like a tool, a template or a workshop seat.
For AEO, that position gets even stronger. If an engine can't read the content behind your form, it can't cite it, quote it or learn from it.
Here's the Bottom Line
AEO isn't a trick. It's the same foundation, with a higher bar. Answer your buyers' real questions first, clearly and honestly. Back it with original data. Show up where your buyers already look. Then measure whether you're named, and how you're described.
I became the HubSpot guy by making tutorials since 2013. Consistency is the superpower nobody talks about.
Here's how I said it on HubHeroes: "If you just focus on like trust in human insights and best practices of structuring and paying attention to the humans that you're actually serving... then it doesn't matter if it's SEO, if it's social, if it's surface."
My advice? "Stop creating single use content." Build an ecosystem where every piece you create sells, ranks, surfaces in AI, gets spread on social and turns customers into supporters. That's the 5S content framework, and AEO lives right in the middle of it.
And full transparency: we're working this playbook on our own site too. We've made sidekickstrategies.com agent-readable with markdown views, an llms.txt file and an MCP server.
Your Next Step
If you want a sidekick to run this with you, from the first-hour audit to the 30-day plan to the weekly review, talk to Sidekick about SEO and AEO. We'll start with the questions your buyers are already asking.



