Why small business websites in Ostrava don't bring in work: what I measured

A small business website is often built once, left untouched for years, and then the owner wonders why it brings in no enquiries. So on 25 October 2026 I measured what the search an AI assistant draws on actually sees of small businesses in Ostrava, and what state their websites are in. For 14 everyday queries such as "plumber Ostrava" it returned 129 results and not one of them was a local business's own website. Directories, job ads and national portals showed up instead.

Don't read this as Google rankings, I did not measure those. It is the view of an AI assistant, which more and more often picks who to recommend. And the conclusion matches what I see with clients: a website or an online shop is not a one-off investment. Without ongoing care and SEO a business cannot be found, and the work goes to whoever is visible, even if they are not better.

What I measured and how

This study follows my technical review of 598 Ostrava business websites from September. This time I did not look at technology but at two business questions: can a person (or an AI) find you by service and does the website convince them to get in touch.

• When: 25 October 2026, a single measurement.
• Queries: 14 local queries " Ostrava" in Czech: plumber, electrician, car repair, accountant, lawyer, dentist, hairdresser, beauty salon, photographer, carpenter, painter and decorator, removals, cleaning company, estate agent.
• Tool: the web search used by an AI assistant (Claude). I recorded the first 10 results and their type: own business website, directory, job portal, other.
• AI answers: for each service a natural-language question "Which … in Ostrava would you recommend?". I recorded the businesses named and the sources cited.
• Business sample: 83 businesses that appeared in the results (68 via a directory, 15 via an AI answer). I found an own website for 45; 34 worked and could be measured.
• Websites: homepage rendered in a browser on desktop and mobile, a 10-item rubric with 0 to 2 points each, Lighthouse in mobile mode, company age from the Czech business register ARES based on the company ID on the website.
• What I did not do: I did not scrape Google, did not download directories (Firmy.cz forbids it in robots.txt), did not submit forms. Results are aggregated only, without business names.

Detailed limitations are at the end. The most important one first: an AI assistant's search is not Google. It has no map pack, does not know the searcher's location and sees small Czech websites poorly. That is exactly why it is interesting. When a customer asks an AI, the answer is assembled from what this search finds.

Finding 1: "service + Ostrava" showed directories, not business websites

Of 129 results for 14 queries, not one was the own website of a specific local business. Most often directories appeared (Firmy.cz 15 times, Českéstavby and Nejřemeslníci 10 times each), job portals with ads like "plumber wanted", and national or unrelated pages.

For 6 of 14 queries directories made up at least half of the results. In the first three places a directory appeared in 16 of 42 cases, an own business website in none.

What it means: when someone asks an AI assistant who can fix a tap in Ostrava, the assistant has nothing about your website to build an answer from. It has a directory. Whoever has no complete profile and reviews there is missing from the answer.

Finding 2: the AI recommended businesses, but based on directories and job ads

For 14 questions "Which … in Ostrava would you recommend?" the AI named 49 businesses in total, a median of 4 per question. The answers relied on 129 sources and not one of them was a business's own website. 64 sources (49.6%) were directories, the rest other pages. 21 sources (16.3%) were job ads and they appeared in 8 of 14 answers. So when recommending a tradesperson, the AI also drew on an ad in which the business was looking for staff.

This is the core of the "weaker competitor" problem. The AI does not choose based on who has the better website or the better work. It chooses from what it finds. If search cannot see your website and you have no directory profile, it recommends a competitor who is there, even if their service is weaker.

How to run the same test for your own field, including a table for recording results, is in my article Does ChatGPT recommend you?.

Finding 3: even a query for the business name did not return its website

For the 45 businesses with a website I also tried the query " Ostrava". The own website was in the top 10 for 0 of 45 businesses. The results were filled with directory profiles: for 79 of 83 businesses at least one directory appeared in the results for their name, for 47 of 83 five or more, the median was 6.

This number needs careful reading. Google would very likely find a business's website for its exact name. This finding says mainly something about the index an AI assistant draws on: small Czech websites are almost absent from it, directories are not. For a business that has a practical consequence. What an AI says about you often comes from your directory profile, not your website. So it pays to keep profiles complete, with the same details as on the website and with reviews. How to do that in Ostrava is in my article Local SEO in Ostrava, and the most common mistakes in Google Business Profile.

Finding 4: a quarter of the websites did not work

Of the 45 websites that directories or AI answers pointed to, 11 (24.4%) did not work or did not load during measurement:

• 6 websites loaded, but showed a hosting message instead of the business. One said the site was "switched off due to unpaid operating costs", four sat on subdomains of the old IC.cz hosting, which now reports "this page is unavailable" instead, and one pointed to a domain not set up at a hosting company.
• 5 websites did not load at all: refused or reset connection (3), an invalid security certificate (1) and a page that kept redirecting itself while loading (1). Some of them may only have refused my crawler, so I count them separately.

For a customer this is the worst possible outcome. They find the business in a directory, click through to the website and see an error or someone else's message. That is exactly what happens when a website is built once and then nobody looks after the domain, hosting and certificate.

Finding 5: working websites sell poorly

The remaining 34 working websites received 0 to 2 points for 10 items that decide whether a visitor understands the offer and gets in touch.

The median was 12 of 20 points, the best website scored 18, the worst 4. Only 4 of 34 websites scored at least 15 points.

I see three things as the most costly:

• The website does not say what and where. Only 11 of 34 websites have both the service and the city in the title or main heading. Yet that is exactly how people search. Without it a website will struggle to appear for "service + Ostrava", however nice it looks.
• The website is ageing. I found a year in the footer on 24 websites, and on 9 of them it was older than 2025, on 7 even 2022 or earlier. Only 16 of 34 websites had the current year or recently dated content. Visitors and search engines read that as nobody looking after the site.
• There is nothing on the website to quote. 7 of 34 websites had a blog, advice section or how-to content, only 9 had a price or price list. AI assistants and search engines need concrete answers (what it costs, how long it takes, what is included). A business card with contact details gives them nothing like that.

I measured speed with Lighthouse in mobile mode on 31 websites (the measurement failed on 3). The median performance score was 54 out of 100, 10 websites scored below 50 and a single one scored 90 or more. The largest content element (LCP) rendered in the lab test in a median of 7.5 s, and in more than 4 s on 26 of 31 websites. This is a single lab measurement with the default network and CPU slowdown Lighthouse simulates for mobile, not real-visitor data, but the direction is clear. The technical results are covered in more detail in the 598 websites study.

A website is not a one-off investment

The six websites that showed a hosting message instead of the business are the clearest illustration: a website does not disappear all at once, but because for years nobody looks after the domain, hosting and content. For 25 working websites I found the company's age in ARES based on its company ID. The median was 15.3 years, so these are established businesses with work and customers. Yet their websites often look as if they were built once and never touched again. Younger and older businesses barely differed in the rubric (median 13 versus 12 points), so it is not about the age of the business but about whether anyone looks after the website.

My own website is how I make my living. Not because it was built once, but because I keep adding services, prices, examples and articles and I watch what people search for to find me. I recommend the same approach to clients. A website needs the same care as a shop: when the offer, prices or opening hours change, the change has to reach the website and the profiles too. Why and how often is in my article Why website maintenance matters, and what it costs in How much website management costs.

How to check where you show up

This study does not tell you where you rank on Google. But you can find that out for free and more precisely than anyone can from outside. Answer two questions: which queries do you show up for and in what position, and does AI cite you.

On points 3 and 4: according to its announcement of 3 June 2026, Google made the generative AI performance reports available to all websites as of 31 August 2026. Microsoft launched AI Performance in Bing Webmaster Tools on 10 February 2026 as a public preview. If you have neither Search Console nor Bing Webmaster Tools, that is the first task. Without them you do not know whether the problem is that nobody finds you, or that people come and do not get in touch.

Search Console also shows clicks for every query. If for your main service and city you see impressions but no clicks, or the query is not in the report at all, enquiries from search are going elsewhere.

What to do first

According to the Firmy.cz help, adding a business to Firmy.cz is free and approval of a new profile takes at least 3 working days. None of this requires a new website. For most of the measured websites it would be enough to add what is missing and start maintaining them. If you do not have the time, I can take it over as part of website management or SEO.

Limitations

• Small, non-random sample. 14 queries, 83 businesses, 45 websites found and 34 measured. The sample includes businesses that already have some online presence (at least a directory profile). Businesses with no presence at all are not in it.
• A single point in time. Everything is from 25 October 2026. Search results change from day to day.
• An AI assistant's search is not Google. No map pack, no searcher location, and an index that sees small Czech websites poorly. I did not measure Google rankings and they cannot be inferred from this study.
• One AI. The answers of one tool are not the same as ChatGPT, Gemini, Perplexity or Google's AI Overviews. Those cannot be measured without logging in and bypassing protections, which is why the how-to above is there.
• Non-working websites. I count as non-working only a website that showed a hosting message or an unconfigured domain instead of content. A website that did not load may only have refused my crawler. It checks the rendered homepage against fixed rules. It can miss a price list hidden in a PDF or references on a subpage. Lighthouse is a single lab measurement in mobile mode.
• Website not found does not mean it does not exist. For 38 of 83 businesses I did not find an own website. Some may have one; the search just did not show it.

Frequently asked questions

Does this mean Ostrava businesses are not on Google?

No. I did not measure Google, and its results look different, with maps and business profiles, for example. The study shows what the search an AI assistant draws on sees. Where you rank on Google is in the Performance report in Search Console.

Why didn't you measure Google directly?

Google does not allow automated retrieval of search results in its robots.txt, and results also differ by the searcher's location. Search Console shows your real queries and positions more accurately than anyone could from outside.

Is a Google profile and a Firmy.cz listing enough without a website?

As a start it is the cheapest step, and according to this measurement the AI really does draw on directories. But a profile cannot carry the details of your services, prices and process. The website is where the decision is completed, and a source an AI can quote concrete answers from.

Which businesses were in the sample?

I do not publish names or domains. The aim is not to point at anyone but to describe what is common. I excluded my own clients, chains and franchises from the sample in advance.

How often should I check Search Console?

Once a month is enough. Watch which queries you show up for, whether positions for your main services are improving and whether clicks are growing. After changing the website or your offer, look sooner.

Can you help me with this?

Yes. I will go through your Search Console, profiles and website, list what is missing and order it by impact. Get in touch via the contact page.

Measured on 25 October 2026. Aggregated data and the methodology are available from me on request, without business names or domains.