AI Agents Are Now Visiting Your Website — Does It Actually Work for Them?

Most Singapore businesses have now heard about AI search — being cited in ChatGPT or Perplexity answers. Agentic traffic is a different, later stage of the same shift, and it is worth separating clearly.

AI citation is an AI system reading your content and mentioning your business in an answer. The user then decides whether to visit you.

Agentic traffic is an AI system visiting your website itself, on the user's behalf — navigating pages, extracting details, filling in forms, comparing your offer against competitors, and in some cases completing an action such as a booking or purchase. The user may never load your site in their own browser at all.

This is no longer theoretical. Both OpenAI and Google have shipped agentic capabilities that let their platforms handle steps across a customer's journey — from discovery through to checkout — with agents able to complete purchases inside the AI interface. Adobe data cited in early 2026 showed AI-referred traffic to US retail sites growing 805% year-over-year on Black Friday 2025, and other analyses put growth in AI-sourced traffic above 1,000% while traditional search traffic declined slightly.

Importantly, this traffic converts. Shoppers arriving from AI services have been found to be around 38% more likely to buy than those arriving through traditional channels — because, as with AI citation generally, they arrive pre-qualified, having already been filtered and matched by the agent.

For a Singapore SME, the near-term question is not "will my customers use AI agents to buy from me." It is: when an agent does arrive on my site to evaluate my business, can it actually understand what I offer and complete what it came to do?

An AI agent experiences your website in a fundamentally different way from a human visitor. A person reads visually — they infer meaning from layout, images, positioning, and context. An agent reads structurally — it parses the underlying code and data, and it struggles with anything conveyed only visually.

This produces several common failure points on otherwise perfectly good Singapore business websites:

Information that exists only in images. Pricing in a graphic, service details in a designed banner, opening hours in a photo of a shopfront sign. A human reads these instantly; an agent frequently cannot. Anything an agent needs must exist as text.

Missing or incomplete structured data. Schema markup is how you tell machines, unambiguously, what your business is, what services you offer, where you operate, and what your key facts are. Its importance for agentic contexts is measurable: research cited in early 2026 found that products with complete Schema.org markup were 6.4 times more likely to be selected by AI agents for recommendations. The same logic applies to services — clear, complete structured data makes you selectable.

Inconsistent or outdated information. An agent comparing options will find contradictions between your website, your Google Business Profile, and directory listings — and inconsistency reduces confidence. Worse, an agent that surfaces stale pricing or an outdated service description creates a customer expectation you then have to correct.

Interactions that require visual navigation. Forms that depend on hover states, multi-step flows with unlabelled fields, or contact paths that only work through a chat widget can all block an agent from completing an action. Notably, the same properties that make a site accessible to screen readers — semantic HTML, proper form labels, keyboard-navigable flows, alt text — are largely the properties that make it usable by agents. Accessibility work and agent-readiness overlap substantially.

Vague, unextractable descriptions. "Empowering businesses through innovative solutions" tells an agent nothing. Specific, factual descriptions of what you do, for whom, in which locations, at what price range, are what allow an agent to match you to a user's request.

The urgency varies by business type, and it is worth being honest about that rather than treating this as universally critical today.

For e-commerce and product businesses, this is closest to immediate. Agentic commerce is already live — protocols and instant-checkout capabilities have shipped across major platforms, and product data completeness, accurate real-time pricing, and structured product information are now competitive infrastructure. If you sell products online in Singapore, auditing your product data and schema is a near-term priority.

For B2B and professional services — the majority of Singapore SMEs — the immediate impact is more about research than transaction. Agents are visiting service websites to compare providers, extract service details, check credentials, and shortlist options. Forrester has predicted that a meaningful share of B2B sellers will encounter agent-led interactions by the end of 2026, and Gartner has projected very large volumes of B2B spending being intermediated by AI agents later this decade. The practical implication now is ensuring an agent evaluating you can accurately extract what you do, who you serve, your credentials, and how to contact you.

For local service businesses, the key surfaces are your website plus your Google Business Profile, with consistent NAP data, accurate hours, clear service areas, and structured local business data — so an agent answering "find me a [service] in [area]" can include you with confidence.

Across all three, the same underlying work applies. The difference is how quickly it becomes commercially consequential.

None of this requires rebuilding your website. The work is mostly structural hygiene, and much of it overlaps with SEO, GEO, and accessibility work you may already be doing:

Test what an agent sees. The simplest starting point: ask ChatGPT or Google's AI Mode to find businesses in your category in Singapore, and see whether you appear and whether the details are correct. Then ask an AI assistant to research your specific business and report back what it finds. The gaps in that answer are your to-do list.

Get your structured data right. Organisation schema, LocalBusiness or Service schema, FAQ schema, and — for products — complete product schema with accurate attributes. This is the single highest-leverage item, because it benefits traditional SEO, AI citation, and agentic contexts simultaneously.

Move critical information out of images and into text. Pricing, service descriptions, hours, contact details, and specifications should all exist as real text on the page.

Ensure consistency across surfaces. Your website, Google Business Profile, and any directory listings should carry identical business information. Contradictions reduce machine confidence in all of them.

Make forms and key actions machine-navigable. Properly labelled form fields, semantic HTML, and flows that do not depend on hover or visual inference. If your site is accessible, it is already most of the way there.

Keep information current. Stale pricing or discontinued services being surfaced by an agent is worse than no information, because it creates expectations you cannot meet.

Do not block agents reflexively. One documented mistake is blanket-blocking agent traffic in robots.txt or via bot protection. There are legitimate reasons to manage automated traffic, but blocking the agents that increasingly mediate customer discovery removes you from that channel entirely. The better approach is to be readable and let the agents work.

A closing note on calibration, because this topic attracts a lot of hype.

Agentic traffic is real and growing quickly, but for most Singapore SMEs it is not yet the dominant source of customers, and it will not be this quarter. The projections that get quoted — McKinsey's estimate of agentic AI influencing trillions in global retail commerce by 2030, Morgan Stanley's expectation that a large share of online shoppers will use AI agents by then — are genuinely large, but they are end-of-decade forecasts, not descriptions of today.

What makes this worth acting on now is not urgency but efficiency: the work that makes your site agent-ready is almost entirely work that already pays for itself in other ways. Structured data improves search visibility and AI citation. Text-based, extractable information improves both SEO and human clarity. Accessible, semantic markup serves screen readers, search engines, and agents alike. Consistent business data across platforms strengthens local search.

In other words, you do not need a separate "agentic strategy" competing for budget. You need the fundamentals done properly — and if they are, you get agent-readiness as a by-product. The Singapore businesses that will be well positioned as this channel matures are simply the ones that did the structural work early, for reasons that were already good on their own terms.

Q1: What is the difference between AI search visibility and agentic traffic?

AI search visibility means an AI system reads your content and mentions your business in an answer — the user then decides whether to visit you. Agentic traffic means an AI system visits your website itself on the user's behalf: navigating pages, extracting details, filling forms, comparing you against competitors, and sometimes completing an action like a booking or purchase, without the user ever loading your site themselves. Both matter, and both depend on similar foundations (clear structure, accurate data, schema markup), but agentic traffic raises an additional requirement: your site must be usable by a machine, not just readable by one.

Q2: How do I check whether AI agents can understand my Singapore business website?

Start with two simple, free tests. First, ask ChatGPT or Google's AI Mode to find businesses in your category in Singapore and see whether you appear and whether the details are accurate. Second, ask an AI assistant to research your specific business and report what it finds — its answer reveals exactly what is extractable and what is missing or wrong. The gaps are your priority list. Beyond that, check whether critical information (pricing, services, hours, contact details) exists as real text rather than only inside images, and validate your schema markup using Google's Rich Results Test.

Q3: Does this only matter for e-commerce, or for service businesses too?

Both, but with different timelines. For e-commerce, agentic commerce is already live — product data completeness, accurate real-time pricing, and structured product information are near-term competitive infrastructure. For B2B and professional services, which make up most Singapore SMEs, the current impact is more about research than transactions: agents are visiting service websites to compare providers, extract details, and shortlist options, with agent-led B2B interactions expected to grow significantly. Either way, the underlying work is the same — accurate, structured, machine-readable information — the difference is how soon it becomes commercially decisive.

Q4: Should I block AI agents from crawling my website?

Generally no, at least not reflexively. Blanket-blocking agent traffic through robots.txt or bot protection is a documented mistake: it removes you from a channel that increasingly mediates how customers discover and evaluate businesses. There can be legitimate reasons to manage automated traffic — server load, scraping concerns, or protecting specific sensitive areas — and those warrant a considered policy rather than a blanket block. For the pages where you want to be found and evaluated (services, products, contact, key content), being readable to agents is the point.

Q5: Do I need a separate budget or strategy for agent readiness?

Usually not, and that is the most useful thing to know about this topic. Almost everything that makes a site agent-ready is work that already pays off elsewhere: schema markup improves search visibility and AI citation; text-based extractable information improves SEO and human clarity; semantic, accessible markup serves screen readers, search engines, and agents alike; consistent business data across your website and Google Business Profile strengthens local search. Rather than funding a separate "agentic strategy," the practical approach is to do these fundamentals properly — agent readiness follows as a by-product, which is why it is worth acting on now even though the channel is still maturing.

Mayson AI builds and optimises Singapore business websites with the structured data, extractable content, and technical foundations that serve human visitors, search engines, and AI agents alike. If you want to see what AI systems currently make of your site, book a consultation.

If you want your website to work for both human buyers and AI browser agents, see Mayson AI's GEO and AI search visibility and SEO website development services.