What actually helps a website earn visibility in AI search

What AI search actually is

AI search is a broad label for products that answer questions by combining language models with information retrieved from the web. Google AI Overviews and AI Mode, ChatGPT search, Perplexity, and Microsoft Copilot do not all find or rank sources in the same way.

Some rely heavily on an existing search index. Some use their own crawler, outside search providers, or a combination of sources. The exact systems change frequently, and none publishes a formula that guarantees a citation.

What they have in common is the need for reliable source material. A useful page has to be discoverable, understandable, relevant to the question, and credible enough to support an answer.

That distinction matters. Answer Engine Optimization, or AEO, is not a separate replacement for SEO. It is a practical way to think about whether your expertise can be found, understood, and used when a person asks an AI-assisted search product a relevant question.

Is traditional SEO still relevant?

Yes. Google's official guidance is explicit that its generative AI features are rooted in the same core Search ranking and quality systems used elsewhere in Google Search.

Google describes two important mechanisms:

• Retrieval-augmented generation: the system retrieves current, relevant pages from its Search index to ground an answer.
• Query fan-out: the system may run several related searches to understand different parts of a broader question.

This means crawlability, indexing, internal links, page quality, relevance, reputation, and technical health still matter. There is no special AEO tag that lets a weak page bypass those fundamentals.

It also means publishing a separate page for every possible wording of a question is the wrong response to query fan-out. A stronger approach is one authoritative page that covers the real topic clearly, including the adjacent questions a buyer naturally needs answered.

See Google's guide to optimizing for generative AI features for its current guidance.

What makes a source useful to AI-assisted search?

The most defensible approach is to create material that is genuinely useful to a person and difficult for a generic summary to replace.

Firsthand experience. Explain what you learned from real work. Include the constraint, the decision, what changed, and where the tradeoff was. A page that reports experience is more distinctive than one that restates common advice.

A clear point of view. Useful expertise includes judgment. If one option is better in certain conditions and worse in others, say so. Avoid vague advice designed to offend no one.

Direct answers supported by evidence. Answer the core question early, then explain the reasoning. Use examples, approved results, dates, screenshots, or other proof when they improve the reader's understanding.

Logical structure. Descriptive headings, short paragraphs, lists, and comparison tables make a complex page easier for people to scan and for machines to interpret. Structure helps, but it does not make weak information authoritative.

Clear authorship and ownership. Identify who created the material, which organization stands behind it, and when it was published or substantially updated. Do not invent expertise signals that the page cannot support.

Useful media. Original images, diagrams, video, and descriptive alt text can provide context that text alone cannot. Generic stock imagery rarely adds the same value.

What do structured data and llms.txt actually do?

Structured data can help a machine understand what a page represents, but it does not make the page more trustworthy by itself and does not guarantee inclusion in an AI answer.

Organization, LocalBusiness, Service, Article, BreadcrumbList, and other schema types are useful when they accurately describe information visible on the page. The markup should agree with the page, not introduce claims that visitors cannot see.

FAQ markup is not an AEO shortcut. In June 2026, Google removed its FAQ rich-result documentation because that feature is no longer shown in Google Search results. A visible FAQ can still be excellent content when it answers real questions, but adding FAQPage markup does not directly earn an AI Overview citation.

The same caution applies to llms.txt. Google says it does not use llms.txt for Search and that the file has no positive or negative effect on Google visibility or rankings. It can remain a lightweight reference for services that choose to use it, but it should not be treated as a core ranking investment.

Structured data is best understood as labeling. The value still comes from the underlying content and the evidence behind it.

Which technical foundations matter?

The technical goal is simple: important public content should be easy to crawl, render, understand, and revisit.

Crawler access. Review robots.txt, page-level robots directives, and CDN or firewall rules. OpenAI says sites that want to appear in ChatGPT search should allow OAI-SearchBot. Crawler access is eligibility, not a placement guarantee.

Indexable pages. Keep valuable public pages indexable and remove utility, duplicate, private, and thin pages from the index where appropriate.

A current sitemap. The XML sitemap should contain the public canonical URLs that matter and update automatically as content changes.

Canonical consistency. Each page should resolve cleanly to its preferred URL. Redirect old URLs intentionally and avoid conflicting canonical signals.

Server-rendered, semantic content. Important copy should be available in the delivered HTML, organized with sensible headings and landmarks. Do not require a fragile interaction before the core answer appears.

Performance and reliability. Fast, stable pages are easier for people and crawlers to retrieve. Core Web Vitals are a useful diagnostic framework, not a special AI citation score.

Accurate metadata. Titles, descriptions, dates, authorship, image metadata, and structured data should describe the page precisely.

These are foundational safeguards. Once they are working, better evidence and reputation usually deserve more attention than another technical micro-optimization.

Why does reputation beyond your website matter?

Your own website can explain what you do. Independent sources help confirm it.

Consistent business information across Google Business Profile, reputable directories, professional profiles, client sites, interviews, awards, and press coverage gives search systems more context about the organization. Natural links and mentions also help connect your claims to evidence outside your own domain.

For a design studio, that may include:

• Detailed client reviews that describe the actual work and outcome
• Client or partner pages that credit and link to the studio's role
• Case studies supported by approved quotes, results, or launch details
• Consistent services, location, name, and contact information across trusted profiles
• Relevant local, industry, or project coverage that mentions the studio accurately

This is not a reason to manufacture reviews or chase low-quality directories. A few specific, credible references are more valuable than a large volume of generic listings.

How should AI search visibility be measured?

Measure business outcomes and source visibility separately. A citation is useful only if it builds meaningful awareness, engagement, or demand.

Search Console. Review Google's generative AI performance reporting when it is available for your property, alongside normal queries, pages, impressions, clicks, and conversions.

Referral analytics. OpenAI says ChatGPT search referral links include utm_source=chatgpt.com, which makes those visits identifiable in analytics. Track AI-assistant referrals as their own channel and inspect which landing pages attract engaged visitors.

Cited-page monitoring. Record which pages are being surfaced or cited for a small, stable set of buyer questions. Results can vary by model, location, personalization, and date, so treat prompt checks as directional rather than a ranking report.

Lead quality. Read the actual inquiry. A form submission is not automatically a qualified lead, and a citation is not automatically revenue.

Change history. Note when an important page, profile, case study, or technical rule changes. AEO results rarely have a clean single-variable explanation, so a basic timeline prevents false conclusions.

There is no responsible universal promise that a page will be cited within a certain number of days. Indexing, competition, query demand, authority, and platform behavior all affect the timeline.

What should businesses avoid?

Do not publish commodity content at scale. Generic pages that summarize what is already online give people and AI systems little reason to choose your source.

Do not create hidden crawler-only copy. Important claims should be useful and visible to visitors. Schema and metadata should reflect the page rather than carrying a second version of it.

Do not build one page for every query variation. Google explicitly warns against scaled content created primarily to manipulate rankings or generative AI responses.

Do not treat FAQ schema, llms.txt, or crawler permissions as ranking hacks. They can support discovery or understanding in limited ways, but they cannot substitute for expertise, proof, or reputation.

Do not promise citations or conversion rates you cannot verify. AI search products change quickly, attribution is incomplete, and small samples can be misleading.

Do not optimize away the human reader. If a page is awkward, repetitive, or unhelpful to a person, its AEO strategy is already off course.

Final thoughts

AI search changes the interface, but it does not remove the need to earn trust.

The durable work is not mysterious: publish firsthand material worth citing, connect claims to proof, keep the site technically clear, build a consistent reputation beyond your own domain, and measure qualified outcomes instead of chasing screenshots of citations.

That work improves traditional search, AI-assisted search, and the experience of the person who eventually reaches your site. It is slower than a shortcut, but it is also much harder for a competitor to copy.