AI Overviews Ranking Factors
The signals that influence whether your content appears in Google AI Overviews — and how to strengthen them.
Cogniva is a Melbourne-based search visibility and AI systems consultancy. We help Victorian businesses earn visibility in Google AI Overviews and across the wider AI search layer. Our approach to this topic is deliberately evidence-weighted: because the published data on AI Overview ranking factors is fast-moving and frequently contradictory, we distinguish between what's strongly supported, what's plausible, and what's hype — and we build strategies on the first category, not the third.
The phrase "ranking factors" is slightly wrong for Google AI Overviews, and the distinction matters before any list makes sense. Google AI Overviews don't rank pages into positions — they run a retrieval-augmented pipeline that pulls passages from indexed pages and synthesises an answer, citing the sources it drew from (the system retrieves relevant passages from indexed pages and uses them to construct the response, rather than generating purely from training data) (SEOcrawl, 2026). So you're not optimising for a rank. You're optimising to be one of the handful of passages Google decides to use — and the factors that influence that are correlations observed across studies, not levers Google has confirmed.
That framing is the whole point of this guide, because the field is awash in confident factor lists built on thin or contradictory evidence. Below, the factors are grouped by how much real evidence supports them. For the strategy above this — whether Google AI Overview visibility is worth pursuing for your business at all — start with our view on where GEO fits a Melbourne business.
First, the honest data problem
Here's the contradiction every other guide steps around. The single most important question — does ranking well on Google get you cited in its AI Overviews? — has genuinely conflicting answers in the published data.
Ahrefs analysed 1.9 million citations in mid-2025 and found 76% of cited URLs ranked in the organic top 10. By February 2026, using a larger dataset, that figure had collapsed to 38% — with the caveat that part of the decline reflects improved measurement methodology, though the directional trend is real (ZipTie, 2026). Meanwhile other 2026 analyses still report that the large majority of AI Overview citations come from top-ranking pages. Both can't be precisely true, and the reason they differ — different datasets, different methods, different months — is exactly why you should distrust any guide quoting one number as gospel.
What survives this mess is a directional conclusion, not a precise one: traditional ranking still helps, probably a lot, but it has weakened as a guarantee, and a meaningful share of citations now come from pages outside the top 10. One reading found 47% of AI Overview citations come from pages ranking below position five. (ZipTie, 2026) So strong SEO remains your foundation, but it's no longer sufficient on its own. Hold that as the frame for everything below.
The factors, ranked by evidence strength
The most useful recent work here is a meta-analysis rather than a single study. In May 2026, Cyrus Shepard of Zyppy synthesised 54 experiments, patents and case studies into a single scored ranking of factors — the first attempt to weight GEO advice by evidence strength rather than opinion (Zyppy meta-analysis via DigitalApplied, 2026). I've used that as the backbone and flagged where individual claims are weaker.
Because this field is full of confident lists built on thin data, every factor below is graded by how much evidence actually stands behind it. Read the grade before you read the factor:
A note on what these grades are not: even "High confidence" here means strong observed correlation, not a Google-confirmed ranking mechanism. Almost nothing in AI Overview optimisation is officially documented — the grades rank the evidence, not Google's intentions.
Crawlability and indexability.
This is the floor, and the most evidence-backed factor of all. In the Zyppy analysis, URL accessibility scored highest — a page typically needs to be crawlable and not blocked, paywalled, or erroring for an AI engine to cite it at all. A surprising number of sites disqualify themselves here without realising. One specific, overlooked trap: pages using a nosnippet directive to suppress preview snippets can inadvertently reduce or eliminate their AI citation visibility, because the engines lean on snippet data during grounding (Zyppy meta-analysis via DigitalApplied, 2026).
Where this model breaks down
In keeping with not overstating a contested field, here's where the factor lists above genuinely wobble.
The headline correlations are correlations, not confirmed mechanisms.
As flagged in the legend, almost none of this is confirmed by Google — these are features consistently correlated with citations across studies, not levers Google has documented (Megrisoft, 2026). Act on the consistent signals; never treat the precise numbers as settled.
The data shifts fast and the studies disagree.
We opened with the 38%-vs-76% contradiction for a reason — it's the rule, not the exception. The honest disclaimer attached to the best meta-analysis is that AI citation behaviour shifts fast, and underlying correlations should be re-checked before treating any single figure as doctrine. (Zyppy meta-analysis via DigitalApplied, 2026) A factor list confident to two decimal places is more suspect than one that admits uncertainty.
Some widely-repeated tactics don't survive scrutiny.
Not everything in circulation holds up. Press-release distribution, a staple of traditional digital PR, accounted for just 0.04% of citations in one large dataset — effectively invisible to AI citation systems. (Zyppy meta-analysis via DigitalApplied, 2026) If a guide recommends it for AI visibility, that guide isn't checking its own data.
Google's index isn't ChatGPT's.
Everything here is Google-specific. The same algorithm update that cut Google AI citations sometimes left Perplexity citations flat or higher, suggesting Perplexity uses a less Google-dependent model. (ZipTie, 2026) Optimising for AI Overviews is not the same as optimising for ChatGPT — which is why our cross-engine checklist separates the actions by engine.
The takeaway isn't paralysis. It's that the defensible strategy is the boring one: fix accessibility, win classic SEO, build genuine brand mentions, structure for extraction, and keep content current — in roughly that order, because that's what the strongest evidence supports, regardless of which month's citation percentage you believe.
What actually breaks AI Overview visibility (the failure patterns)
A list of what works is only half a diagnosis. In the audits Cogniva runs for Melbourne businesses, pages usually fail to get cited not because they're missing an exotic factor, but because of a handful of recurring, fixable problems. Framing these as Cogniva's observed patterns rather than published statistics — these are professional judgements from auditing real sites, not a study — here's what we see most.
The page is generic — it says what everyone else says.
This is the most common one by a distance. Google's pipeline has little reason to cite a passage that merely restates what three other sources already said more clearly. If your content adds no information the model can't get elsewhere, it loses the selection step even when it's technically well-optimised. The fix is information gain: a genuine fact, figure, angle, or piece of first-hand experience the competing pages don't have.
The business has no entity grounding.
Plenty of pages are well-written but describe a business the wider web barely corroborates — thin or inconsistent listings, no recognisable author, a brand Google's Knowledge Graph can't confidently identify. The page can be perfect and still lose, because the model can't verify who's behind it. This is usually an off-page problem masquerading as an on-page one.
The answer is buried.
A page that does contain the answer, three scrolls down, after the brand story and the preamble, often won't be cited — the extractable answer has to be near the surface of the relevant section. We see strong content lose to weaker content purely on structure.
Internal cannibalisation.
When a business has four overlapping pages all loosely targeting the same query, none of them is the clear, authoritative answer — they split the topical signal between them. Consolidating into one comprehensive page frequently does more for citation than any amount of per-page tweaking.
Cogniva's perspective:
if you're going to spend on AI Overview visibility, diagnosing which of these is happening to your pages is worth more than working down the factor list blind. Three of the four failures above are invisible in a normal rankings report — which is exactly why most businesses never find them.
Why this is worth the effort (the click data)
The reason any of this matters commercially: citation increasingly beats ranking for actual clicks. When AI Overviews appear, organic click-through on those queries drops sharply — one large study measured a 61% fall. (ZipTie, 2026) But the flip side rewards the cited: pages cited inside an AI Overview earn around 35% more organic clicks and 91% more paid clicks than uncited competitors. (Search Engine Land via Mike Khorev, 2026) The traffic doesn't vanish — it concentrates on whoever Google chose to cite. In a market where most businesses haven't adjusted to this, that's a real opening — the broader case for it is in our GEO Melbourne guide.
The strategic read is straightforward: as answers consolidate onto a handful of cited sources, visibility stops being a gentle slope and becomes closer to a short list you're either on or off. Being one of the cited few is worth disproportionately more than being the eleventh blue link ever was — which is why the effort to earn citations, while the field is still thin, pays back well above its cost.
Good to know
Frequently asked questions
Straight answers before you book anything — and if yours isn’t here, ask us in the consultation.
Free consultation
Are Google’s AI Overviews sending you traffic — or taking it?
We’ll analyse which of your queries trigger AI Overviews and whether your content is positioned to be cited in them.
Book a consultation