Brands are no longer competing only for a blue link. They are competing to become part of the answer.
Marketing leaders want to know why ChatGPT recommends one provider, Perplexity cites another and Google AI Overviews assembles its response from sources that may not hold the traditional number-one position.
The market has answered with a flood of Generative Engine Optimization advice, much of it built on speculative tactics and rebranded SEO basics. Our experience points to a less fashionable but far more useful conclusion: AI search optimization works when a brand becomes easy to retrieve, understand, verify and recommend.
That requires strong SEO, clear entity signals, genuinely useful content and independent evidence that the market recognises the brand beyond its own website.
These patterns should not be treated as a universal scoring system. They show where work has built genuine authority and where activity has often created noise without improving the case for recommendation.
AI Search Builds Defensible Answers
ChatGPT, Perplexity and Google AI Overviews do not work identically, but their search experiences share an important characteristic: they retrieve and synthesise information from web sources. ChatGPT Search can reformulate a question into targeted searches and return source links. Perplexity describes its answers as grounded in current web sources with inline citations. Google says AI Overviews and AI Mode can use query fan-out, running related searches across subtopics before assembling a response with supporting links. A recommendation can therefore depend on several layers of evidence:- Whether your pages can be retrieved
- Whether the system understands what your company does
- Whether your content addresses the decision being made
- Whether independent sources support your claims
- Whether the brand fits the user’s market, use case and constraints
Citation, Mention and Recommendation Are Different
AI visibility is often treated as one metric. It is not. A citation means your page was used to support an answer. A mention means the brand appeared in the response. A recommendation means it was presented as a viable choice, usually in relation to a specific requirement. A company can earn citations without being recommended. Its research may help explain a market while its own commercial proposition remains unclear. A brand may also be recommended because several third-party sources describe it consistently, even when its website is not directly cited. Citation growth requires retrievable, useful source material. Recommendation growth also requires entity clarity, commercial relevance and external corroboration. Across Verta’s optimization work, the strongest results have come from improving the entire evidence environment around a brand rather than chasing isolated AI citations.What Our Observations Show About AI Visibility
These are observed campaign patterns, not officially confirmed ranking factors. No platform publishes a complete formula for AI citations or recommendations, and no credible agency should pretend otherwise.Authority Pages With Commercial Substance
Many websites have hundreds of articles and almost nothing substantial on the pages explaining what the business sells. That is backwards. Strong service, solution, and product pages establish the facts an AI system needs: who the offer is for, which problem it solves, what makes it different, where it is available, what it integrates with, and what evidence supports its claims. The best authority pages combine commercial clarity with real expertise. A cybersecurity company should explain its coverage, deployment model, supported environments, and relevant use cases. A software platform should make its capabilities, limitations, integrations, and ideal customer profile explicit. Supporting content should deepen the same subject area. This is topical authority in a useful sense: not publishing every conceivable keyword variation, but building a coherent body of information around problems the company is qualified to solve. Google’s guidance for generative AI search similarly prioritises unique, expert-led, non-commodity content over pages that simply restate what already exists.Unambiguous Brand Entities
Weak entity signals create avoidable uncertainty. A company may describe itself differently across its website, LinkedIn profile, directory listings, and partner pages. Product names change without explanation. Author pages contain no credentials. The About page tells a brand story but never clearly states the company’s category, markets, or expertise. Entity SEO makes these relationships explicit. Use consistent company names, product terminology, positioning, and factual business information across credible platforms. Build meaningful organisation and author pages. Clarify the relationships between the company, its products, experts, locations, and industries. This also means developing a consistent semantic identity. A business cannot describe itself as a consultancy on one platform, a software provider on another, and a full-service agency elsewhere, then expect search systems to infer its primary market position correctly. Structured data can support this clarity. Google says Organization markup can help it understand and disambiguate an organisation, but it also states that no special schema is required for its generative AI features. Schema should confirm reality, not manufacture authority.Third-Party Corroboration
A company website is a self-published claim. Recommendations are much easier to defend when credible third parties corroborate, validate, or provide context for that claim. Industry publications, editorial coverage, partner directories, expert roundups, podcasts, review platforms and independent case studies can help establish that a company belongs within a recognised commercial ecosystem. The quality of the association matters more than mention volume. Ten generic directory profiles do not equal one detailed inclusion in a respected industry resource. A syndicated press release is not independent recognition. A backlink from an unrelated high-authority website may improve a third-party SEO metric without making the brand more relevant to the recommendation being generated. The strongest digital PR connects a brand to a specific expertise area, product category, result, methodology or point of view. You cannot self-declare your way into market authority.Decision-Support Content
Recommendation prompts usually contain criteria and trade-offs: “Which platforms are best for a multi-location clinic?” “What are the strongest alternatives to this software?” “Which agency suits a B2B SaaS company entering the US?” The most useful source material helps answer the decision. We have repeatedly seen value in well-built comparison pages, alternatives content, buyer’s guides, implementation frameworks, use-case pages and category explainers. The objective is not to publish a self-serving “best” list with the company conveniently placed first. It is to explain who each option serves, where it performs well, which limitations matter and when another choice may be better. This degree of honesty can feel commercially uncomfortable. It is also what makes the page credible enough to cite. Google confirms that its AI features may issue multiple related queries across a topic. OpenAI similarly explains that ChatGPT Search can reformulate questions into targeted searches. Depth across real decision criteria is therefore more useful than mechanically targeting one exact phrase.Original Information Worth Reusing
Commodity content gives an AI system little reason to select one source over another. Original research, first-party data, expert commentary, named frameworks and detailed implementation lessons create information that can travel. They can earn links, citations, discussion and independent references because they add something to the market. This does not require an enterprise research budget. A consultancy can analyse patterns across completed audits. A software company can publish anonymised product usage trends. A specialist can document a decision framework developed through years of client work. A manufacturer can explain how specifications affect performance in actual operating conditions. The advantage is not simply fresh content. It is source-worthy knowledge.What Actually Moved the Needle
| Signals That Helped | Signals With Little Observable Impact Alone |
| Deep authority around a defined commercial subject | High volumes of generic blog content |
| Clear company, product and expert entities | Keyword and brand-term repetition |
| Credible third-party mentions and editorial links | Low-quality directory and press-release saturation |
| Detailed comparison and decision-support content | Thin “best tools” articles created only to target prompts |
| Original research, frameworks and expert insight | AI-generated summaries without expert contribution |
| Strong service pages connected through internal links | Informational content with no commercial architecture |
| Clean crawlability and indexation | Treating schema as an AI ranking hack |
| Consistent brand facts across reliable sources | Chasing isolated GEO shortcuts |
AI Search Optimization Still Starts With SEO
The fastest way to waste money on generative engine optimization is to treat it as a separate channel while the underlying website remains weak. For businesses asking how to rank in AI Overviews, Google’s own guidance is direct: pages must be indexed and eligible for Search before they can appear as supporting links in AI Overviews or AI Mode. OpenAI recommends allowing OAI-SearchBot so content can surface in ChatGPT Search, while Perplexity provides similar guidance for PerplexityBot. Important pages still need to be:- Crawlable and indexable
- Internally connected
- Canonicalised correctly
- Free from unnecessary duplication
- Accessible without rendering barriers
- Organised within a clear site architecture
A Practical Programme for Better AI Visibility
The patterns themselves are not particularly surprising. What matters is how consistently they appear across industries, platforms and search experiences. AI visibility has introduced new surfaces, but not entirely new rules. Most successful projects end up strengthening the same underlying signals of expertise, trust and relevance. The question is how to prioritise those signals in practice.1. Define Commercial Prompts
Do not begin with hundreds of vanity prompts. Build a focused set around category discovery, comparisons, alternatives, use cases, integration requirements and provider recommendations. Include the qualifiers buyers actually use, such as industry, geography, company size, budget, compliance requirements and existing technology.2. Audit the Evidence Behind Each Answer
Review which brands and sources appear for important prompts. Look beyond the final response and examine what makes the answer defensible. Then assess your own footprint. Is there a page that directly supports inclusion? Is the brand category clear? Are credible third parties reinforcing the same positioning? Is the available information accurate and current?3. Strengthen Owned Authority
Improve service, product, industry and use-case pages before scaling another broad editorial calendar. Create supporting content where it adds expertise, resolves a buyer objection or helps the reader make a decision. Publishing more content is not a strategy when the website’s commercial core remains vague.4. Build Independent Recognition
Use digital PR, expert contributions, partnerships, reviews and industry participation to create accurate third-party evidence. The objective is not indiscriminate mention volume. It is a consistent association with the categories, capabilities and problems the business wants to own.5. Measure a Portfolio, Not a Screenshot
AI answers vary by wording, model, timing, location and search behaviour. One favourable prompt result proves very little. Track:- Citation share
- Brand mention rate
- Recommendation rate
- Source diversity
- Competitor inclusion
- Prompts where the brand is repeatedly absent
- Referral traffic from AI-powered search
