Whether you're looking for new ideas to boost performance, impress your boss/client, or mix it up from the same old projects you run every month, the list below will help. I have been in marketing for 10+ years, and I just got back from Unbound, HubSpot's annual marketing conference. One of my favorite things to do is learn new actionable items from industry experts.
And with the AI era we have entered, a list like this has never been more important. So many fundamental marketing practices have changed, and having something that starts to map out the new industry can be make or break.
I have decided to write a few pieces with all my findings from the conference. The one you're reading now focuses on the new findings around AEO, or AI visibility: how to show up more in AI searches.
One thing I want to call out before the list: nearly every item below is backed by real experiment data, not general, unproven "best practices." These are findings the experts tested and measured, and I have provided both the stats and the source so you know they're worth your time before you invest any.
Let's start with some of the big shifts that surprised me most (my favorites).
If you want AI to quote you, write the format it quotes most. Knowing which formats it pulls from is probably the single most important thing on this list for generating traffic in 2026.
|
Rank |
Type |
% of AI Citations |
Citations vs a Regular Webpage |
|
1 |
Best-of / Listicle |
32% |
3x more often |
|
2 |
How-to Guide |
14% |
2.7x more |
|
3 |
Original Research / Data Study |
12% |
2.6x more |
|
4 |
Statistical Roundup |
9% |
2.3x more |
|
5 |
Comparison / "vs" Content |
8% |
2.2x more |
|
6 |
Step-by-Step Tutorial |
7% |
2.1x more |
|
7 |
Checklist |
5% |
1.9x more |
|
8 |
Case Study |
4% |
1.8x more |
|
9 |
Template / Framework |
4% |
1.7x more |
|
10 |
FAQ |
3% |
1.6x more |
(Source: Jay Schwedelson / Worldata Research.)
What surprised me: Best-of and listicle content is clearly the most valuable, and it's something we have been under-utilizing. It's cited more than 2X as often as the next format on the list, and 3x more often than an average page.
For a while, the whole industry was excited about LLMs.txt, a plain-text file you add to your site as a "map" for AI models, the way robots.txt works for search crawlers. It made intuitive sense. Everyone rushed to build one.
Then HubSpot actually tested it on their own site and found zero bots visiting the LLMs.txt path. Not "underperforming." Zero. That's what put this on my radar, and the independent research backs it up:
Why it matters: This is the most useful kind of finding, the one that saves you time rather than adds to your list. If you're a lean team, building an LLMs.txt file is effort with no return right now. The bigger lesson the HubSpot team drew from it is worth keeping: don't assume you know the result before you test it, and don't chase industry hype without checking. (Source: HubSpot's AEO experiments, presented by Amanda Kopen, Manager of AI Search and Interactive Experiences at HubSpot.)
One of the highest-ROI AEO assets is also one of the least glamorous: a deep, structured glossary that defines every core term and concept in your category, all in one place. When someone, or an AI, asks "what does X mean," a clean glossary entry is the exact, extractable answer the model wants to grab.
Why it matters: When HubSpot built one, the glossary drove a 36% lift in visibility where they were cited, the content outperformed their other content by 15%, and it re-ranked for hundreds of keywords along the way. Better still, unlike most technical AEO work, a marketer can build a glossary without a developer. If there are foundational concepts your category owns (or that you want to own), that's where to start. (Source: HubSpot's AEO experiments, presented by Amanda Kopen.)
Instead of guessing what your buyers ask, ask the AI directly. The method, shared by Jay Schwedelson, is to prompt each major model (ChatGPT, Gemini, Claude) to generate the questions real prospects ask about your category, then look at the questions that overlap across all three. That overlap is your content plan, sourced straight from the tools your buyers are actually using.
I tried Jay's original version and found it needed some tweaks. So I rebuilt it to stay strictly at the category and industry level, avoid including my brand or company name, and aim more at the audience actually doing the research. Here's the refined version. Swap the bracketed parts for your brand and audience:
You are analyzing how real people use ChatGPT, Gemini, Claude, and other LLMs to research topics and make decisions in this category. Review this brand's website, [BRAND NAME], [BRAND URL], along with its products, pricing, competitors, and marketing materials, to understand the category and industry the brand operates in. [Add 1-2 sentences describing the strategy/category in plain terms: the specific approach, the factors involved, and the relevant category topics it falls into.] Do not generate any questions about [YOUR BRAND NAME] or any specific named product or brand. I want only category-level and industry-level questions, the general topics this brand falls into, not the product itself. Generate the 50 questions most likely to be asked of an LLM by real [TARGET AUDIENCE, e.g. Registered Investment Advisors] researching these industry topics. Factor in the types of questions you commonly receive about this category, the strategies and problems within it, competitor approaches, and common patterns in how [that audience] uses AI to learn and evaluate ideas. Prioritize by likelihood of actually being asked, not by what would be useful or flattering for any brand. Include discovery, education, comparisons of approaches, pricing and fee questions, alternatives, proof and evidence, problems, objections, risks, and implementation. Write each question exactly as a real [audience member] would ask it. The goal is to surface topics they will find organically based on these industry and category angles, so keep every question at the topic level and free of any product name. Please provide in a numbered list.
Why it matters: It replaces guesswork with the actual questions buyers type into AI, and the overlap across models is your highest-confidence roadmap. Every piece you make from it is aimed at a question someone is already asking. (Method credit: Jay Schwedelson.)
For twenty-plus years, backlinks were the dominant currency of getting found. More links, higher rankings. When HubSpot looked at what predicts AI visibility, that relationship basically disappeared. Pages with huge backlink profiles weren't consistently cited more, and some frequently-cited sources had almost no link authority at all.
Why it matters: Google looks for the most authoritative page. Answer engines look for the best snippet to answer the question. So clarity and specificity often beat domain authority, which means a smaller brand can compete on being the clearest answer rather than the biggest name. If you've been measuring your AEO progress by your old SEO scoreboard, that's the first habit to drop. (Source: HubSpot AEO research, presented by Aja Frost, Senior Director of Global Growth at HubSpot.)
Answer engines are literally trying to answer questions, so make it easy for them. Turn your page headers into the exact question a buyer would type ("What is AEO?" instead of "AEO Definition"; "How do I do AEO?" instead of "AEO Best Practices"). Check out all of the headers in this blog…
Why it matters: When your page structure already mirrors the question, it's dramatically easier for the model to extract your content and reuse it as the answer. It also happens to be better for humans, who scan for their exact question too. (Source: Aja Frost, Senior Director of Global Growth at HubSpot.)
Answer engines lean harder on recency than Google does. A few concrete moves that came up: put a visible "last updated" date on your pages, put a recent date or time period right in the title ("The State of X, Q3 2026"), drop timestamps into the copy itself ("as of August 2026"), and genuinely refresh old content rather than just changing the date.
Why it matters: A visible last-updated date correlates with more citations, and putting a recent date in the title showed a 50%+ higher likelihood of appearing in ChatGPT and other LLMs. Recency tells the model the information is still safe to repeat. And refreshing old content is cheap, it lets a proven page re-enter the citation pool without starting from scratch. (Sources: independent data from AIROps & Ahrefs; refresh tactics from Jay Schwedelson.)
Certain content types get dramatically more downloads in the October-through-January planning window, when your buyers are deciding what next year looks like:
Why it matters: A next-year industry calendar saw a 128% lift in downloads in that window. Same content, published at the moment your audience is actively planning, performs on a different level. One honest caveat for my own world: the Q4 timing reflects general B2B behavior, so if your audience is on a different cycle, gut-check it against your own data. (Source: Jay Schwedelson.)
AI treats specific, unique data as more credible and more quotable. Original benchmarks, case-study results, and statistics that no one else has are exactly what makes a page "snippable," and they're the one thing a competitor can't copy off you.
Why it matters: When a model is deciding what to cite, concrete data does triple duty: it's specific, it's trustworthy, and it's easy to quote. You don't need a giant research budget either, your own customer outcomes and internal numbers are data no one else can publish. (Source: Aja Frost, Senior Director of Global Growth at HubSpot.)
We are guilty of this one, and it was a punch to the gut to hear, considering how much we have placed in images over the years.
If your key message or important stats live inside a JPG or PNG, AI can't read it. Crawlers skip right over text trapped in a graphic, so a critical point stuck in a designed image simply doesn't exist as far as the model is concerned.
Why it matters: This is a free fix. Anything you want AI to see, headlines, key points, the actual substance, has to be live text on the page, not a picture of text. It's the same reason your dev team should care about JavaScript (more on that below). (Source: consistent theme across the Unbound AEO sessions.)
Structured data (schema markup) has been sold as an AEO growth lever. The data doesn't back that up. Ahrefs tracked 1,885 pages that added JSON-LD schema and found no citation boost across AI platforms.
Why it matters: This reconciles a genuine debate, some speakers swear by schema, others say it does nothing. The honest read is that schema helps on Google's AI surfaces but does little for ChatGPT, so it's basic hygiene worth doing, not a lever that will move your numbers on its own. Keep adding it. Just don't expect it to be the thing that gets you cited, and don't let anyone sell it to you that way. (Source: independent data from Ahrefs.)
AI agents are increasingly visiting sites on behalf of users, and they won't recommend what they can't parse. That means the practical facts, pricing or fees, specs, availability, policies, need to be in plain, machine-readable form, not locked in a PDF or an image or behind an interaction.
Why it matters: Agent traffic is projected to keep climbing, and an agent that can't read your fundamentals just moves on to a competitor it can read. Getting the basics machine-readable now is getting ahead of where this is going. (Source: Marcus Sheridan.)
AI crawlers are far less capable than Google's twenty-year-old crawler. They don't run JavaScript well, they choke on heavy pages, and when a page is slow or unclear they often just leave. Two fixes came up repeatedly:
Why it matters: If your content is invisible when JavaScript is off, or your pages load slowly, you may be invisible to AI without knowing it. This is a conversation to have with your technical team, and the question to open with is simple: can we check our crawl logs and see how fast AI bots are actually retrieving our pages? (Source: HubSpot AEO experiments, presented by Amanda Kopen.)
Most teams track whether they got cited or mentioned. The leading indicator is earlier than that: how often the AI bots are crawling your site at all. If ChatGPT's crawler isn't visiting, citations are impossible, and you'll wait weeks staring at the wrong metric before you realize the problem is upstream.
Why it matters: Crawl frequency tells you whether AI is even reaching your pages before you wait on citations to move. It's the missing metric on most AEO scorecards, and it turns "why aren't we showing up" from a guess into something you can actually see. (Source: HubSpot AEO experiments, presented by Amanda Kopen.)
AI search is volatile and it's early, which is exactly why it's worth moving now. The old SEO scoreboard matters less than being the clearest, freshest, most extractable answer to the specific question your buyer is asking. Structure for extraction, keep it current, put real data on the page, and measure whether the bots are even showing up.
None of this is finished, the whole field is being figured out in real time. But that's the opportunity. The teams that get comfortable with the new rules now will be the ones AI is quoting when everyone else is still wondering where their traffic went.
I love writing content like this. I try to write the kind of thing that, if I stumbled across it in my own LinkedIn feed, would actually give me something to try that affects my everyday work, and this piece is the epitome of that. If it helped you, I'd genuinely like to hear about it. Like it, leave a comment, and follow me on LinkedIn for the rest of the series.