Answer Engine Optimisation for SMEs: Why Traffic Is Falling but Enquiries Are Flat

If your analytics show organic sessions sliding month after month while your enquiry count has barely moved, your website is probably not broken. Your customers are still finding you — they are just reading the answer instead of clicking through to get it. AI answer panels in Google, plus ChatGPT, Gemini, Perplexity and Copilot, now sit between the question and your site. They summarise your page, credit you in a line of small text, and send far fewer people onward. The correct response is not more blog posts and not a panicked SEO retainer. It is answer engine optimisation: making sure that when an AI system composes the answer, the facts it uses are yours, attributed to you, and accurate.

Why is my website traffic falling while my enquiries stay flat?

Before you diagnose anything, segment your pages by type. In almost every case we look at, the loss is concentrated in informational content — "what is", "how much does", "types of", "requirements for" pages. Those are exactly the queries an AI panel can resolve in three sentences. Meanwhile your commercial pages — the ones with your product names, service areas, price lists and contact form — hold steady or grow. That split is the signature of AI interception rather than a penalty or a technical fault.

Check Google Search Console for the same pattern from the other direction: impressions holding or rising while clicks fall. Rising impressions with a collapsing click-through rate means you are being surfaced and read, not demoted. It also means the enquiries you do receive are arriving better informed, further down the buying process, and less willing to wait for a reply — which is why this issue lands on the sales desk before it lands on the marketing budget.

What is answer engine optimisation, and how is it different from SEO?

Classic SEO optimises for a position in a ranked list of links. The click was the product. Answer engine optimisation optimises to be the source a language model retrieves, trusts and cites while writing a single composed answer. There is no position ten to crawl up from; either your facts are in the answer or they are not.

That changes what "good content" means. Ranked search rewarded length, keyword coverage and backlinks. Answer engines reward extractability — a self-contained passage that answers one question without needing the rest of the page — plus unambiguous identity, verifiable specifics, and corroboration from sources other than you. A 2,000-word thought-leadership piece with the answer buried in paragraph nine performs badly. A 250-word page that states a lead time, a service area and a price band in the opening lines performs well.

How do AI search engines decide which Singapore business to name?

Practically, three things decide it. First, retrievability: the content has to be in crawlable HTML. Specifications locked inside PDFs, prices rendered as images, and pages that only assemble themselves after JavaScript runs are invisible to a meaningful share of retrieval systems. Second, entity clarity: the model needs to be confident that the company on your website is the same company in the directory listing, the association member list and the Google Business Profile. Inconsistent business names, three different phone numbers and a missing UEN all dilute that confidence. Third, specificity: models prefer text that commits to something checkable — minimum order quantity, response window, districts served, certifications held, and what you explicitly do not do. Vague positioning language gives a summariser nothing to quote.

What should a 15-person SME change on its website this quarter?

1. Mine your own inbox. List the ten questions your sales or support inbox answers most often. Those are your page titles, in the customer's wording, not your industry's.
2. Answer in the first fifty words. One page, one question, answer at the top. Background, credentials and story go below.
3. Commit to specifics. Publish lead times, coverage areas, MOQs, price bands, opening hours and exclusions. If a number cannot be published, publish the range and the condition.
4. Make the facts machine-readable. Add Organization, LocalBusiness and FAQPage structured data. Keep name, address, phone and UEN identical across the website, Google Business Profile and every directory.
5. Get the facts out of PDFs. Every spec sheet and price list needs an HTML equivalent on a real URL.
6. Date everything and prune. Show a last-updated date, and delete or redirect pages carrying obsolete prices and discontinued products.
7. Build corroboration. Industry association listings, supplier and partner pages, trade directories and a handful of genuine third-party mentions. Answer engines triangulate; a single self-published claim is weak evidence.
8. Decide your crawler policy deliberately. Check whether your robots.txt, CDN or firewall is blocking AI crawlers. Blocking them is a legitimate choice, but it should be a decision someone made, not a default your web developer left in place.

How do you measure answer engine optimisation with no ranking report?

Change the primary metric. Sessions are no longer a proxy for demand, so promote enquiries per week and first-response time to the top of the dashboard and let traffic become a supporting number.

Then run a monthly citation check. Fix a list of twelve to fifteen questions a real buyer would ask, ask them in ChatGPT, Gemini, Perplexity and Google's AI panel, and record in a spreadsheet whether your company is named, which page was cited, and whether the facts quoted were correct. It takes about thirty minutes a month and it is the closest thing to a rank tracker this channel has. Alongside that, watch your server or analytics referrers for AI assistant domains, keep the Search Console impressions-versus-clicks view, and add a "how did you find us?" field to your enquiry form. Owners are consistently surprised how often the answer is now "ChatGPT mentioned you".

What is the risk if this works?

Two things follow success, and both are operational rather than marketing problems. An answer engine will quote your page verbatim, including a stale price or an expired lead time, and it will keep doing so for weeks. Answer engine optimisation is therefore a content accuracy programme with a named owner and a quarterly review, not a one-off project. And the enquiries that arrive will already know your price band and your capability, so they expect a same-hour reply. Winning the citation and then taking four hours to quote simply moves the loss from the traffic report to the order book.

Where should you start if you only have one day?

Publish the five questions your inbox answers most, each as its own page with the answer in the opening lines and at least one hard number. Fix your business name, address, phone and UEN so they match everywhere. Add basic structured data. Put the monthly citation check in the calendar as a recurring thirty-minute task. That is a day's work, it requires no redesign, and it puts you ahead of most of your competitors, who are still buying keyword reports for clicks that are no longer coming.

Frequently asked questions

Should I block AI crawlers to stop them using my content?

Only if being cited has no commercial value to you — for example, if your content is the product you sell. For a services or supply business, the citation is free distribution to a buyer who is mid-decision. Blocking removes you from the answer without restoring the old click. Decide it explicitly, document it, and review it yearly.

Do I need to rebuild my website for answer engine optimisation?

Usually not. The common blockers are content-shaped, not platform-shaped: answers buried too low, facts trapped in PDFs, inconsistent business details and missing structured data. A rebuild is only warranted if your pages render entirely through JavaScript or you cannot edit page content without a developer.

How long before this shows up in enquiries?

Expect two to four months before citation rates move meaningfully, because retrieval indexes and model training cycles both lag your publishing. The faster return is internal: the same pages that make you citable also shorten your quotation and enquiry-handling time, because your team stops retyping the same answers by hand.