7 Findings From 100 Cape Town Business Websites: The Hidden Cost of Cheap and AI-Built Sites

When an audit starts with a traffic decline, use the September 2026 spam-update diagnosis guide to connect the observed symptom with the next useful check before changing successful pages.

If you are evaluating a new website, compare it against the AI-built website SEO failure checklist.

Direct answer: What did the Cape Town website audit find?

Easy Rank audited 100 eligible business websites found through local service searches connected to six Cape Town suburbs. Thirty-five had no detectable JSON-LD schema in the server-delivered homepage. Ninety did not have detectable LocalBusiness schema. Ninety-five did not have Service schema. Fifty-two did not have exactly one H1 heading. Twenty-two had no detectable meta description and 11 had no canonical link.

These findings do not mean that every affected website is absent from Google. They show missed opportunities to communicate business identity, services and page structure more clearly.

AI Overview-ready summary

• Easy Rank audited 221 candidate domains and selected 100 eligible Cape Town business websites.
• The sample included Durbanville, Hout Bay, Mitchells Plain, Bellville, Woodstock and Brackenfell.
• Thirty-five websites had no detectable JSON-LD schema in the server-delivered homepage.
• Only 10 used detectable LocalBusiness schema and only five used Service schema.
• Fifty-two websites did not have exactly one H1 heading.
• Twenty-two lacked a detectable meta description and 11 lacked a canonical link.
• One eligible site was positively identified as Lovable. That single example is not enough to judge all vibe-coded websites.
• The central problem is not the tool used to build the site. It is whether the finished site includes the technical, content and entity layers required for search visibility.

How the study was conducted

Easy Rank conducted the audit on 3 September 2026. We used local search combinations covering plumbers, electricians, accountants, dentists, builders, beauty salons, attorneys and restaurants. Each query included one of the following areas:

• Durbanville
• Hout Bay
• Mitchells Plain
• Bellville
• Woodstock
• Brackenfell

The discovery process produced 221 unique candidate domains. We removed directories, social profiles, government pages, job platforms and websites that did not visibly mention the relevant suburb. Of the 112 eligible sites remaining, 100 were selected through a suburb-based rotation.

The audit inspected the server-delivered homepage HTML for JSON-LD schema, schema types, title elements, meta descriptions, H1 counts, canonical links and indexability signals. We also used detectable code signatures to group platforms such as WordPress, Wix, Squarespace, Framer, Lovable and React-based sites.

Important limitations

This is a search-result sample, not a random census of every Cape Town business website. A site appeared because it was returned for one of the research queries. Inclusion does not certify that the company has a physical address in that suburb.

The audit did not execute JavaScript. Structured data injected only after client-side rendering may therefore be missed. Platform detection is based on visible code signatures and does not prove who built the website or whether AI was used. The suburb sample sizes were also different, so the suburb figures should not be used as a ranking table.

Finding 1: 35 websites had no detectable schema

Sixty-five of the 100 websites had at least one detectable JSON-LD block. Thirty-five had none in the server-delivered homepage.

Schema markup is structured data that describes the meaning of visible information. Google says it uses structured data to understand page content and gather information about entities such as organisations, people, products and places. It can also make pages eligible for certain rich results. It does not guarantee a ranking, rich result or AI citation. Google structured-data introduction

The absence of schema does not automatically make a website invisible. It means the business is relying more heavily on Google and other systems to infer its identity and relationships from ordinary page content.

Finding 2: Only 10 websites used LocalBusiness schema

Only 10 of the 100 audited homepages contained a detectable LocalBusiness type or a recognised local-business subtype.

That is the most commercially important finding. A local company may display its name, telephone number and address to visitors but fail to express those details in a consistent machine-readable business entity.

Google says LocalBusiness structured data can describe information such as business hours, departments and other business details. Eligible information may contribute to business features in Search or Maps, although display is not guaranteed. Google LocalBusiness documentation

For a dentist in Bellville, salon in Woodstock or plumber in Brackenfell, the correct local-business subtype can help define what the organisation is and where it operates. The markup must match the visible page and real business information.

Finding 3: Only five websites described services with Service schema

Five homepages contained detectable Service schema. This does not mean every service page needs a schema block or that Service schema produces a special Google result. It shows that most sites in the sample did not use structured data to connect their organisation with the services described on the page.

A strong implementation can connect the organisation, service, provider, service area and relevant page. The content must still explain the service clearly. Schema cannot repair thin copy, unclear positioning or a missing location page.

Finding 4: 52 websites did not have exactly one H1

Only 48 audited homepages had exactly one H1 heading in the delivered HTML. The remaining sites had no H1 or multiple H1 elements.

An H1 is not a magic ranking switch. It gives readers and search systems a clear primary heading for the page. A homepage built mainly around visual sections, animated components or generic slogans can look impressive while failing to state what the business actually does and where it operates.

The primary heading should describe the page accurately. “Welcome to our website” is less useful than “Emergency Plumbing Services in Durbanville and Cape Town.”

Finding 5: 22 websites lacked a detectable meta description

Seventy-eight websites had a detectable meta description. Twenty-two did not.

Google may rewrite a meta description, so adding one does not control every search snippet. A relevant description still gives the business an opportunity to summarise the page, service and location. It also supports better sharing and quality control across a site.

Cheap website packages often focus on visible layout because clients can approve colours, photographs and sections. Metadata is less visible, which makes it easier to omit during a rushed launch.

Finding 6: 11 websites lacked a canonical link

Eighty-nine homepages included a detectable canonical link. Eleven did not.

A canonical indicates the preferred URL for a page. This becomes important when websites can be accessed through several URL versions, tracking parameters or duplicated routes. Its absence is not automatically a serious fault on every homepage, but it is part of the technical production layer that should be checked before launch.

Finding 7: The platform did not determine quality

WordPress powered 69 websites in the final sample. Fifty-three of those had detectable schema. The remaining platform groups were much smaller, so comparisons require caution.

The sample included five Wix sites, two Squarespace sites, four detectable React or Vite single-page applications, one Framer site and one Lovable site. All five Wix sites and both Squarespace sites had some detectable schema. Three of the four React or Vite sites did not. The Framer site and the single Lovable site had no detectable schema.

One Lovable website is not a study of Lovable and cannot support a general claim about vibe coding. The useful conclusion is broader: no platform removes the need for production QA. WordPress can be poorly configured. A custom site can omit basic metadata. An AI-built site can look polished while its search layer remains unfinished.

What we found across the six suburbs

Mitchells Plain produced only five eligible sites under the strict inclusion rules. This is a finding about the sample and discovery method, not proof that Mitchells Plain has fewer businesses or weaker websites. The figures are useful for identifying patterns, but they should not be treated as a competition between suburbs.

The wider result matters more: 90 of the 100 sites lacked detectable LocalBusiness markup, including businesses associated with established commercial areas and fast-growing northern suburbs.

What this says about cheap and vibe-coded websites

Low-cost and AI-assisted website tools have made design more accessible. That is valuable. A small business can publish a professional-looking site faster than before.

The risk appears when “published” is mistaken for “finished.” A production-ready business website needs more than a layout:

1. Crawlable and indexable pages
2. Clear titles and headings
3. Useful service and location content
4. Accurate metadata and canonical URLs
5. Connected Organisation, LocalBusiness and page-level schema where appropriate
6. Consistent details across the website and Google Business Profile
7. Analytics, Search Console and conversion tracking

Vibe coding is not the problem. Incomplete implementation is the problem. A cheap website becomes expensive when the business later pays to rebuild missing service pages, repair technical SEO or explain conflicting business information across platforms.

A practical seven-step website check

1. Test indexability

Confirm that important pages are not blocked by `noindex`, robots rules, login screens or broken rendering.

2. Check the delivered HTML

Do not inspect only what appears visually in a browser. Check whether titles, headings, canonical links and meaningful content exist in the delivered or reliably rendered page.

3. Validate schema

Use Google’s Rich Results Test and the Schema.org validator. Fix errors, but also check whether valid code accurately matches visible content. Google requires structured data to represent the page truthfully. Google structured-data guidelines

4. Connect business entities

Use stable identifiers to connect the organisation, website, founder, location, services and relevant profiles. Do not invent awards, ratings or service areas.

5. Improve local content

A business serving Hout Bay needs more than the words “Hout Bay” in a footer. Explain the service, customer problem, operating area, process, proof and next step.

6. Compare the website with the Google Business Profile

Check the name, category, address, telephone number, hours, service areas and website link. Conflicting information creates avoidable uncertainty. Compare the website with the Google Business Profile

7. Measure after launch

Connect Search Console, GA4 and conversion tracking. Monitor indexation, search visibility, engagement and qualified enquiries without promising rankings or revenue.