A study analyzing 540,000 query pairs from Ahrefs surfaced a number every SEO should read twice: Google's own two AI search surfaces — AI Overviews (the summary atop results) and AI Mode (the standalone conversational search) — cite the same URLs only 13.7% of the time when answering the same question. Even looking only at each side's top three citations, overlap is just 16.3%. In other words, same Google, same question, and 87% of the cited sources are two entirely different sets of sites.
Why is this only surfacing now? Because over the past 12 months Google quietly split "search" into two systems. AI Overviews now reaches over 2.5 billion monthly active users across 200+ countries and 40 languages; AI Mode crossed 1 billion monthly actives about a year after launch. Both carry the Google name and their answers are 86% semantically similar (89% of pairs score above 0.8), yet each sources differently under the hood: AI Mode uses "query fan-out," splitting one question into several sub-searches then synthesizing, so it leans toward encyclopedic, definitional sources — Wikipedia appears in 28.9% of AI Mode responses versus 18.1% of AI Overviews.
Set against the wider GEO category, the direction is clear: tools that monitor "whether AI mentions you" are exploding — Profound, Ahrefs Brand Radar, Semrush, Scrunch are all fighting for it — but most track only one or two surfaces. When Google itself splits into two citation engines, and ChatGPT and Perplexity each have their own logic, the credibility of any "single score" dilutes fast.
For SMBs and agencies, the meaning is blunt: the decade-old belief that "rank #1 on Google and AI will cite you" is officially dead. This piece answers where to put your effort now that Google itself is two searches — and which parts need no SaaS at all.
What the Numbers Say
Ahrefs compared 730,000 response pairs (semantic similarity) and 540,000 query pairs (cited URLs). The core finding: the two surfaces "say almost the same thing, cite almost entirely different sources." Word-level overlap is just 16%, meaning even the phrasing diverges. Another key backdrop from separate research: only 37.9% of URLs cited in AI Overviews also rank in the organic top 10, down from about 76% in July 2025. In one year, the link between "ranking" and "being cited" nearly halved. This is not a tuning tweak — it is a structural split of ranking and sourcing into two separate jobs.
Immediate Actions for Three Reader Types
- Brand owners / SMBs: Stop watching only "am I ranked first." Test it: take your 10 most important questions, ask each in both AI Overviews and AI Mode, and log who each cites. You will watch yourself appear on one side and vanish on the other.
- Marketers / SEOs: Reshape content into independently quotable blocks — under each subhead, answer the question in one sentence, with numbers, definitions and steps in clean chunks. Both surfaces prefer content they can lift directly; this beats keyword stuffing.
- Developers / agencies: Add structured data (FAQPage, HowTo, Article schema), clear entity markup and the semantic HTML recommended by Google Search Central. That is the foundation both surfaces can read.
AI Visibility Tool Comparison
| Tool | Focus | Price tier | Limitation |
|---|---|---|---|
| Profound | Cross ChatGPT/Perplexity/AIO visibility | Enterprise, high monthly | Pricey for SMBs; uneven AI Mode coverage |
| Ahrefs Brand Radar | AI citations + classic SEO in one | Bundled with Ahrefs | Requires existing Ahrefs stack |
| Semrush AI Toolkit | AI mentions & sentiment | Add-on module | Coverage skews to US market |
| Scrunch | AI search visibility monitoring | Mid-high | Weak Traditional-Chinese / Taiwan data |
What They Won't Tell You
- An "AI visibility score" is usually one surface, sampled: If Google's two surfaces already share only 13% of citations, any dashboard compressing that into one number hides huge variance. A pretty score does not mean you appear on the surface your customer actually uses.
- Chasing "citation overlap" is a trap: Rather than trying to win both at once, build content both would independently pick — clear facts, directly quotable, unambiguous entities. Single-surface black-hat tricks get flattened by the next update.
The No-SaaS SMB Alternative
You do not need an enterprise tool first. DIY path: add FAQPage / Article structured data, write key content as quotable "one question, one conclusion" blocks, and make sure primary text is readable without JavaScript. For monitoring, use one Google Sheet: each week manually ask your 10 most important questions in both AI Overviews and AI Mode and log the cited domains. Two months reveals the trend. Once the process proves its value, scale with a paid tool.
FAQ
What exactly is the difference between AI Overviews and AI Mode?
AI Overviews is the AI summary atop a normal Google search; AI Mode is a standalone conversational search surface. Their conclusions are 86% semantically similar, but citations overlap only 13.7% — you must manage them as two mindsets.
So should I still do traditional SEO?
Yes, but it downgrades from "guarantees citation" to "necessary but not sufficient." Only 37.9% of AIO-cited URLs also rank top 10, so a good ranking is the entry ticket, not the seat.
Must an SMB buy an AI visibility tool?
Not necessarily. Build a baseline with free methods (manual Q&A + structured data + a Google Sheet), prove traffic or inquiry value, then decide on paid. Tools solve scale, not existence.
Do I need separate content for each surface?
No. One piece that is factually clear, directly quotable and fully entity-marked is the common denominator both surfaces favor independently — more durable than customizing per surface.
My Take
The mainstream is busy chasing "how to win AI Overviews and AI Mode at once," but my call is the opposite: this split is a phase, and over the next 12–18 months Google will likely converge its two sourcing engines. So spending heavily now to customize for a single surface is investing in a seam that will close. What resists change is content that is fact-dense, machine-quotable and entity-clear — however many surfaces the future splits into, that content gets picked. The ScriptWalker implication: rather than sell a "guaranteed AI ranking" pitch (no one can guarantee it), position the service as an "AEO-ready website foundation" — structured data, semantic HTML, quotable content architecture. That is a verifiable, recurring, algorithm-agnostic productized offering.
Sources
- Ahrefs, "AI Overviews vs AI Mode" study (730K responses / 540K queries): link (primary)
- Google official blog, AI search scale data: link (primary)
- Google Search Central structured data docs: link (primary)
- Search Engine Journal, analysis of the Ahrefs report: link (third-party)
- SE Ranking, AI Mode sources & volatility research: link (third-party)