Google's AI Overviews work largely as an extension of search, so ranking helps there. AI assistants such as ChatGPT and Gemini select sources in a different way, and ranking has little effect on them. A budget decision starts with knowing which surface is in question.
GEO and SEO serve different machines
The data shows two distinct surfaces. Ahrefs studied 15,000 long-tail prompts and found that, on average, 12% of the URLs cited by AI assistants also appear in Google's top 10 for the same prompt. About 80% do not rank anywhere in Google for that query (Ahrefs, August 2025, updated May 2026). The 12% is an average across five engines and includes Perplexity at 28.6%. Without Perplexity, the other assistants sit nearer 8%.
A separate Ahrefs study covered 863,000 keywords and 4 million AI Overview URLs. It found that 38% of AI Overview citations rank in Google's top 10 (reported by Search Engine Journal, March 2026).
The two studies use different datasets and measure different products. Read together, they show that the surfaces diverge. A budget should follow the surface the buyers use.
Ranking supports AI Overviews, and assistants need topic and entity work
Exhibit 1 sets out how each surface relates to search ranking.
Exhibit 1: The two-surface table
| Google AI Overviews | AI assistants (ChatGPT, Gemini, Copilot) | Perplexity | |
|---|---|---|---|
| Share of citations that rank in Google's top 10 | 38% (Ahrefs), around 17% (BrightEdge) | Roughly 6% to 9% | 28.6% |
| Behaves like | An extension of search | A separate retrieval system | A hybrid |
| Does ranking help? | Yes, materially | A little | Yes |
| What to fund | Conventional SEO | Topic and entity work | Both |
A note on the figures. The AI Overview figures and the assistant figures come from two Ahrefs studies with different datasets and methods. Ahrefs also states that its earlier 76% AI Overview figure and the current 38% are not directly comparable, because its citation parsing improved between the two studies. Part of the change reflects better detection. Ahrefs publishes two of the assistant figures without naming the engine each belongs to, so this article reports both as roughly 8%.
Perplexity sits apart from the other assistants at 28.6%, consistent with a product designed to cite its sources.
A first place ranking on Google earns citations on one surface. Assistants reward topic depth and a clear entity.
AI engines search around a topic, beyond the typed keyword
Google documents this behaviour directly. Its Search Central documentation states that AI Overviews and AI Mode may use a "query fan-out" technique. The engine issues several related searches across subtopics and data sources to build a response (Google Search Central).
Fan-out changes how a keyword strategy performs. Ahrefs gives an example. A page that ranks around sixth for "how to descale a coffee machine", "cleaning a Nespresso machine" and "remove limescale from coffee maker" may be cited ahead of a page that ranks first for only one of them. Consistent presence across the cluster earns the citation. Ahrefs also suggests that assistants may sometimes treat the user's prompt as a starting point for other queries. The data is consistent with this possibility, and the mechanism is not confirmed.
Reciprocal Rank Fusion (RRF) is a merging method often used to explain why consistently appearing pages win. Ahrefs describes citations as merged using methods "like" RRF and sources this to a third-party blog. Search Engine Journal notes that Google has not confirmed specific changes to fan-out behaviour. Fan-out is documented by Google. RRF is an industry inference.
How to get your brand mentioned by ChatGPT explains how answer engines select citations. This article covers what to fund.
The topic is now the unit of optimisation
Fan-out moves the unit of optimisation from the keyword to the topic. When an engine breaks a prompt into related questions, a brand needs pages that cover the whole cluster. Keywords still matter, as part of a topic.
Exhibit 2: Keyword, topic, entity. The three units of optimisation
| Unit | What it is | Rewarded by | The work |
|---|---|---|---|
| Keyword | One phrase, one page, one rank | Classic search, and AI Overviews in part | On-page SEO |
| Topic | A cluster of related phrasings the engine fans out into | AI assistants, through query variation. Fan-out is documented for Google's AI features. Fusion methods are an industry inference | Depth and coverage across the cluster |
| Entity | What the engine believes you are | Every surface. Vantage's working model, not a documented mechanism | Consistency across owned, declared, earned and inferred sources |
The entity row describes Vantage's working model of what these systems reward. Google and OpenAI have not documented it as a mechanism. In this model, an entity is the engine's working belief about what an organisation does, whom it serves and what it is credible for. The engine builds this belief from four kinds of source: owned, declared, earned and inferred. A brand with a clear entity enters the keyword contest with an advantage. How to make sure AI describes your brand correctly covers entity work in full.
Retail offers a useful comparison. Ranking resembles shelf position in one aisle. The entity resembles the shop the customer has already chosen to trust. Ahrefs reaches a similar view. Its closing advice is that optimising for query variations, building topic clusters and "owning the entity" "seems to be the way to go". The evidence points in this direction without proving it.
Fund both disciplines, in proportion to how buyers search
The useful question is which surface the buyer uses, and what that surface rewards.
For buyers who search Google and see AI Overviews, conventional SEO continues to pay. Google's documentation states that SEO best practices remain relevant for its AI features, with no additional requirements to appear in AI Overviews or AI Mode.
For buyers who ask an assistant, topic depth and entity consistency carry the brand. Most considered B2B brands in this region serve both groups. A procurement lead often searches Google for a shortlist, and a founder often asks ChatGPT for one.
Google's guidance adds two points for planning. First, no new machine-readable files, AI text files or special schema are needed to appear in these features. An llms.txt file and a schema block are therefore supporting measures. Second, Google treats separate content created for every search variation, including fan-out queries, primarily to influence rankings or generative responses, as scaled content abuse under its spam policy. Fan-out rewards genuine depth. Both disciplines draw on one budget, so plan the split around the buyer's behaviour.
What this looks like in Southeast Asia
Regional brands need a plan for each language. A regional B2B brand often optimises English keywords for Singapore while its buyers in Indonesia and Vietnam ask assistants in their own languages. Where an engine fans a prompt out, the sub-queries are likely to follow the language of the prompt. A brand with one strong English page may then appear in none of them. This is an inference from documented behaviour, not a measured finding.
The Ahrefs dataset includes a Spanish-language example prompt, so the study covered more than English. No Southeast Asia-specific citation statistic was found at source, and this article offers none.
A brand operating in four markets runs four topic clusters in four languages. The engine fans out in the language the buyer uses. The entity layer stays global: the engine holds one belief about what the organisation is, built from sources in every market. Consistent positioning across the Singapore site and the Indonesian site gives the engine one clear entity. Why Southeast Asian brands struggle to scale covers the wider regional pattern.
Measure share of the answer alongside rank
AI search calls for a new measure. Rank tracking describes one surface well. Share of the answer, the proportion of relevant AI answers that cite the brand, describes the others. How to measure brand visibility in AI search sets out the method, and how AI is changing branding covers the wider shift.