Back in early 2024, a client of mine, a mid-size legal services firm in Manchester, watched their organic traffic drop 31% in six weeks without losing a single ranking position. They were still number two for their main terms. The AI Overview box had simply... eaten their clicks. And the site getting cited inside that box? A smaller competitor with a domain authority of 28, no backlinks worth mentioning, and content published four months prior.
That moment changed how I think about search. Because it proved something I now tell every client: Google's AI Overviews do not reward who ranks highest. They reward who answers best. Those are two very different briefs.
So let's get into what actually gets you cited.
---
AI Overviews Are Not Just a SERP Feature
Most people are treating AIO (AI Overviews) like a rich snippet or a featured snippet with better PR. Wrong framing entirely.
Featured snippets pulled a verbatim block of text from a page. AI Overviews synthesise across multiple sources and cite the ones that contributed the most useful, specific, defensible information. Google's own documentation describes them as being powered by a customised version of Gemini, built to ground responses in web content. Which means the selection logic is more like RAG (retrieval-augmented generation) than traditional ranking.
The practical upshot: your page can sit at position six, have moderate authority, and still get cited, if it contains the right kind of content in the right structure. I've seen this repeatedly across the 12,000+ sites we've worked on at Seahawk. The correlation between rank position and AIO citation is real but it's much weaker than people assume. Position one gets cited more often, yes. But it's nowhere near guaranteed.
---
The Content Signal That Actually Moves the Needle
Specificity Over Comprehensiveness
Here's something I tell my content team constantly: a page that answers one question precisely will outperform a page that answers twenty questions vaguely. Every single time, when it comes to AI citations.
Think about why. The AI is trying to construct a trustworthy, accurate answer for a user. It needs source material that is specific enough to quote. A page that says "there are several ways to improve your conversion rate, including design, copy, and trust signals" gives the AI nothing to work with. A page that says "changing the CTA button colour from grey to orange on our client's checkout page lifted conversions by 14% in a 30-day A/B test using Google Optimize", now we're talking.
Specificity is the unlock. Numbers. Named tools. Dates. Real outcomes. The more your content reads like a case study or a primary source, the more useful it is to an AI trying to ground its answer in something credible.
At Seahawk we had a fintech client last year where we rewrote three blog posts to replace vague best-practice language with actual data from their own platform. Two of those three posts started appearing in AI Overviews within eight weeks. The third didn't, we later figured out it was because the topic (regulatory compliance) was one Google treats as a YMYL area and applies extra scrutiny to the source.
Structure: Headers Are Not Optional
The AI needs to parse your content. Google's Search Central documentation has long emphasised structured data, but for AIO the structural signal goes deeper than schema.
Your H2s and H3s need to be proper questions or direct answers. Not clever. Not creative. Direct. If someone searches "how long does it take for a WordPress plugin update to break a site", your H2 should probably say something close to that question verbatim, not "Plugin Updates and Their Risks" which is the kind of heading that belongs in a 2014 whitepaper.
Short paragraphs matter too. Blocks of 200-word prose are harder for a language model to extract a clean citation from. I aim for paragraph lengths of 60-90 words for any section I actually want cited.
---
E-E-A-T Is Not a Checkbox, It's a Pattern
Google's E-E-A-T framework (Experience, Expertise, Authoritativeness, Trustworthiness) gets mentioned in every SEO article in 2024 as if ticking four boxes on a checklist will do it. It won't.
What E-E-A-T actually means for AIO citation is: does your page demonstrate that a real, qualified person produced this based on first-hand experience? Not "our team of experts" in the author bio. Actual signals.
Here's how I think about it in practice:
- Author pages with verifiable credentials. A named author, a real LinkedIn profile, a consistent byline history. Not "Staff Writer".
- First-person experiential language. "When I tested this on a WooCommerce 8.3 install..." beats "According to best practices..."
- Original data. Even small surveys. Even your own analytics. Something the AI can't find anywhere else.
- Publication dates and update dates. Freshness matters enormously for AIO. An article last updated 14 months ago is a liability on fast-moving topics.
I added proper author schema to about 40 client sites in Q3 2023 using Yoast SEO's author structured data feature. Across those sites, we saw a measurable uptick in AIO appearances on informational queries within three months. Correlation, not causation, I know. But it fits the pattern.
---
Topics Where AIO Rarely Cites You (Save Your Energy)
Not every topic is worth optimising for AI Overview citation. Worth being honest about that.
YMYL topics (health, finance, law) get cited almost exclusively from sites with institutional authority. NHS.uk, GOV.uk, major financial regulators. If you're a smaller health blog, you're probably not getting cited no matter how good your content is. That's not a content quality problem, it's a trust classification problem.
Highly commercial queries. "Best project management software 2024" type queries. AIO tends to cite aggregator sites or avoid the format entirely because the answer is inherently contested. You might rank fine but the box may not even appear.
Breaking news and rapidly changing topics. The AI struggles to cite content when the facts are still in flux. Your freshly published article on a story from two days ago is unlikely to get pulled.
Where AIO loves to cite: how-to procedural content, definitions, comparisons with concrete criteria, factual explanations. These are your targets.
---
Technical Factors People Are Ignoring
Page Speed and Crawlability
This one gets hand-waved. It shouldn't. If Googlebot can't efficiently crawl and render your page, the content isn't going in the index cleanly, which means the AI doesn't have clean material to work from.
I ran a crawl audit on a SaaS client's blog in February using Screaming Frog and found 34 pages with render-blocking scripts that were causing Googlebot to time out before fully rendering the page content. These were all informational posts. Zero of them appeared in any AIO result at the time. Fixed the render issues, resubmitted via Search Console, and within six weeks four of those pages had AIO citations.
Coincidence? Maybe. But I've seen it too many times now to dismiss it.
Clean, Parseable HTML
The AI is processing your page content. Tables with merged cells, JavaScript-rendered content that doesn't fall back gracefully, dynamic tabs that hide half your text, all of this makes it harder for the AI to extract clean, coherent information.
Write content in plain, hierarchical HTML where you can. If you're on WordPress (which most of our clients are), make sure your block editor output isn't wrapping content in div soup. The Classic Editor or a well-configured Gutenberg setup is fine. Some of the more baroque page builders are genuinely making your content harder to parse.
---
A Practical Workflow for Targeting AIO Citation
This is roughly what I do now when I want a specific page to have a fighting chance at being cited.
- Identify the exact query format. Search the topic yourself, see if AIO appears. If it does, look at what's being cited. That's your competition.
- Write a dedicated H2 that mirrors the query intent exactly. Not cleverly. Directly.
- Put your best, most specific answer in the first 100 words of that section. Don't bury the lede.
- Add concrete specifics. A number, a named tool, a date, a real outcome. Something the AI can extract and attribute.
- Add proper author schema. Yoast, Rank Math, or manual JSON-LD. Get a real person's name on the content.
- Check your Core Web Vitals. LCP under 2.5s, no layout shift above 0.1. Use PageSpeed Insights, it takes four minutes.
- Update the published date only when the content genuinely changes. Fake freshness is easy to detect. Real freshness signals matter.
- Build one or two contextually relevant internal links to the page. Not a footer link. A link from a thematically related post with anchor text that reflects the topic.
---
What Won't Work (And I've Tried)
Adding "AI Overview optimised" schema that doesn't exist. Keyword stuffing the target query into the meta description hoping the AI picks it up. Publishing thin content fast to capture a trend. Buying links to boost authority specifically for AIO (links matter for general authority, but a sudden spike won't move AIO citation in the short term).
The thing that consistently fails is trying to game the model rather than serve the reader. AIO is a reader-proxy. It's asking: would a well-informed human trust this page as a source? If yes, cite it. If the page feels like it was written for an algorithm, it reads that way to the AI too.
---
FAQ
Does ranking position still matter for AI Overview citations?
It matters, but less than most SEOs think. Pages ranking between positions 3 and 10 regularly get cited if their content is more specific and better structured than the top result. Position one still has an edge, but it's not deterministic.
Can I tell if my site has been cited in an AI Overview?
Sort of. Google Search Console doesn't yet give a clean AIO-specific report, though this is changing. The practical workaround is to search your target queries manually (in a private window, logged out) and check. There are also third-party rank trackers like SE Ranking and Semrush that are building AIO tracking into their toolsets, though coverage is still patchy as of mid-2024.
Should I change my content strategy entirely to chase AIO citations?
No. AIO citation should be a layer on top of a fundamentally sound content strategy, not a replacement for it. Pages that get cited are almost always pages that were already doing the basics well: clear structure, real expertise, useful specifics. If those aren't in place, chasing AIO is putting the cart before the horse.
Does schema markup directly influence AIO citation?
Not directly, as far as anyone can verify. But author schema, article schema, and FAQ schema all make your content more parseable and more trustworthy-looking. Think of schema as raising the floor rather than raising the ceiling.
---
The honest summary: AI Overviews favour pages that are specific, structured, credibly authored, and technically clean. None of that is new advice. What is new is that the penalty for ignoring it is now immediate and visible in the form of a box that sits above your listing and sends the reader somewhere else.
That Manchester legal firm, by the way? We rewrote four of their core informational pages in March. By May, two of them were getting cited. Traffic didn't fully recover, clicks from AIO are just different now, but they stopped watching a competitor benefit from their own ranking effort.
That felt like a win worth having.