I attended a run of AirOps webinars over the past year, Jairo Guerrero, Chris Long, Eli Schwartz, Mike King, Lashay Lewis, Josh Grant, Dave Gerhardt, and a few others. I'm consolidating what I learned here as a resource for clients working on their own sites.
Ranking and Getting Cited Are Two Different Jobs
An AI system takes one prompt and fans it out into a dozen sub-questions, retrieves broadly across the web, then narrows down to a small set it actually cites. An AirOps study of over 548,000 pages found that only about 15 percent of retrieved pages made it into the final answer. Ranking well in Google doesn't guarantee any of that traffic. Track citation rate and mention rate as their own metrics, separate from search rank.
Crawlability Comes Before Content
Most AI platforms don't crawl a site directly. They pull a snippet from Google or Bing's existing index first, and only send their own bot when that snippet falls short. Page speed and uptime affect whether that bot shows up. A lot of this activity doesn't appear in Google Search Console, so server logs are the only way to catch pages an AI bot never visited. Most models also read raw HTML and don't execute JavaScript, so any button, form, or interactive element built as a styled div instead of real semantic HTML is invisible to them.
Structure Every Page for Extraction
Models extract individual passages, not full pages. Put the direct answer near the top instead of building to it. Use action-led step headings with a bridge sentence between each one. Add a comparison table anywhere a real tradeoff exists, comparison tables lifted citation likelihood by 47 percent on commercial-intent pages in one AirOps study. This applies just as much to comparison and alternatives pages, which carry extra weight because an AI system usually can't answer a "best," "versus," or "alternatives" question from its own training data. It has to retrieve and cite an outside source, so open those pages by naming the buyer's persona and situation before naming the product category, and pull the real language from sales calls and support tickets instead of guessing at it.
Citations and Mentions Are Not the Same Metric
A citation is when an AI answer pulls from and links to your page. A mention is when your brand gets named with no link at all. Citations can be engineered through content structure. Mentions come from off-site presence, reviews, forums, PR, people talking about you where you didn't ask them to, and they're harder to manufacture. Track both, separately, and build reporting around three tiers: performance (referral traffic, conversions), channel (citation rate, sentiment), and input (bot activity, page speed).
Off-Site Presence Is Doing More Than You Think
Long-tail, high-intent content and off-site authority matter more under AI search than they did under traditional search, since prompts run longer and more specific than a typical query. Authentic participation in places like Reddit shapes citation and sentiment. Self-promotional posting doesn't. AEO also doesn't have to sit only in an SEO team's dashboard, Webflow's own results showed real signup conversions coming directly from AI search traffic, which makes the case for treating it as a growth channel with an owner, not a defensive checklist.
Keep It Fresh and Refresh on a Schedule
AI systems favor content that's current and backed by original data over generic, templated pages. Refresh timestamps, statistics, and examples on your highest-value pages on a set cadence. Comparison and alternatives pages specifically need a six to twelve month refresh, since competitor content in that category changes fast and stale claims undercut a model's confidence once it's already cited you.
A Practical Checklist
Content structure
- Audit your highest-traffic, highest-intent pages for a direct answer within the first few sentences, not buried after paragraphs of setup.
- Use action-led step headings on how-to content, with a transition sentence connecting each step.
- Write each paragraph so it can stand alone as an answer. Don't rely on the reader having read the paragraph above it.
- Add a comparison table anywhere a real tradeoff exists between options.
Comparison and alternatives pages
- Publish or refresh these before broader use-case content. They tend to get pulled into AI answers more than general informational pages.
- Open by naming the buyer's situation and problem before naming your product or category.
- Pull the actual language buyers use from sales calls and support tickets rather than guessing at it.
- Put these pages on a six to twelve month refresh cycle. Competitor comparison content changes fast, and stale claims undercut trust once a model has cited you.
Technical access
- Confirm no conversion-critical content depends on client-side JavaScript to render. Most AI systems read raw HTML and don't execute scripts.
- Build interactive elements, buttons, forms, filters, as real semantic HTML, not styled elements that only look clickable.
- Check page speed and uptime on your priority pages. Both affect whether an AI bot bothers to crawl at all.
- Pull server logs periodically to see whether AI bots are actually reaching your key pages. This won't show up in Google Search Console.
Measurement
- Separate citation tracking (your content pulled and linked in an answer) from mention tracking (your brand named with no link).
- Report across three tiers instead of relying on traffic alone: performance (referral traffic, conversions), channel (citation rate, sentiment), and input (bot activity, page speed).
- Pick a small set of grounded, high-intent prompts to track continuously, rather than a broad list of generic ones.
Off-site presence
- Show up authentically in the communities where your buyers actually talk, Reddit threads, industry forums, review sites. Self-promotional posting backfires; genuine participation doesn't.
- Treat video, YouTube especially, as part of your retrieval footprint, not a separate channel.
- Keep your positioning and terminology consistent everywhere your brand appears. Inconsistency makes it harder for a model to build a confident picture of what you actually do.
This is treated as an ongoing workflow in every one of these sessions, not a one-time audit. AI answers can shift week to week even when the underlying page hasn't changed, so the review needs a regular place on the calendar.
If you want a second set of eyes on where your site stands with any of this, get in touch.
