Many online shops have an agency or one person who "looks after the shop". Orders come in, the site loads, nobody complains. It is easy to conclude that everything is fine. A working shop, however, does not mean it is doing what it could. Technical state, SEO, visibility in AI, catalogue structure and product feeds can be neglected for a long time without orders showing it right away.
This article is not about how to choose an agency. That is covered in How to choose an SEO agency and How to choose a web agency. Here I deal with the next step: you already have someone and want to know whether they are really looking after the shop.
- Orders are not enough. They say the shop works, not that it makes use of organic search, comparison sites and AI assistants.
- Check what can be measured: how many products are in the feeds compared with the catalogue, how the shop looks on mobile, whether cookies match consent, what titles and descriptions look like and whether the shop appears in AI answers.
- Ask for evidence: a list of changes from the last three months, access to your own accounts and how the agency measures success.
- A second opinion is the most reliable check when it comes from someone who does not manage the shop and therefore is not grading their own work.
- Price: an audit of a Shoptet shop with me costs €280 and I deduct it from follow-up work if you decide to order it within 60 days. An SEO audit of a website starts at €400; an e-shop outside Shoptet is quoted by scope.
Why the owner does not see the problem
The reasons are simple, and nobody needs bad intentions to cause them:
- The shop can be used. The customer picks goods, pays and the order arrives. Errors in a comparison-site feed or in product titles do not produce any error window.
- Reports show other numbers. A monthly report often contains traffic and a few positions. How many products reached the comparison site or how many categories have export switched off may not be in it.
- Nobody independent measures it. Whoever manages the shop also grades their own work. That does not mean dishonesty, only that a second pair of eyes is missing. Just as bookkeeping is checked by someone other than the person who keeps it.
- Errors build up gradually. A template change, a new supplier, a data import. Each looks harmless on its own; together, a year later, they add up to a state nobody would want to see at first glance.
So this is not about someone neglecting something. The point is different: the owner did not know because nobody independent measured it.
What I found on two online shops
Two examples from my practice, both anonymous: no names, domains, industries, towns or supplier names. I give figures as shares and ranges. I do not mean to criticise anyone; I want to show what a state the owner does not know about looks like.
- Shop A has a catalogue of over 10,000 products. I measured it in September and October 2026 from feeds, exports and the administration. The findings come from my first measurements in September and October 2026 and do not say who caused them or when.
- Shop B has a few hundred products. On 27 October 2026 I audited it from the public website, without access to the administration.
| Area | What I found | Why it costs money |
|---|---|---|
| Feeds (A) | The feed for a goods comparison site received less than 1% of the catalogue, because in more than 9 of 10 categories export was switched off in the administration. | Products that are not on the comparison site cannot be ordered through it. |
| Feeds (A) | In the Google feed, less than 1% of items had a product category and more than a quarter had a wrong availability status (pre-order instead of "in stock at supplier"). | Google categorises products wrongly and shows customers untrue availability. |
| Feeds (A) | Dozens of sale prices higher than the regular price and dozens of invalid EANs. | Systems reject or flag such items as faulty. |
| SEO (A) | About two thirds of product names ended with a year or a season. For some, removing it would create duplicate names, so it was handled carefully. | A needlessly long product name that goes out of date quickly. |
| SEO (A) | SEO titles were assembled automatically by the system in the form "name only for price CZK", longer than what shows in results, with a price that changes. | The title is truncated in search results and the price in it goes stale. |
| Catalogue structure (A) | Descriptions for a large brand were patchworks from the supplier's feed: article code, untranslated English, the same text again in Czech. About every eighth short description was in English only. | The customer does not know what they are buying and the search engine has nothing to read. |
| Technical state (B) | On mobile, the subcategories of the main menu were unreadable (dark text on a dark background). | A customer on a phone cannot see where to go. |
| Technical state (B) | Four images on the homepage did not load: the link returned a web page instead of an image. | A homepage with empty spaces looks unfinished. |
| Technical state (B) | In tests, an advertising cookie was stored before the customer consented and stayed after they declined. | Legal risk and unreliable measurement. |
| Technical state (B) | Mobile homepage: lab LCP of about 8 s, over 8 MB transferred (much of it a video in the intro). Within the limit on desktop. | A slow first page on a phone. |
| Catalogue structure (B) | For about half of the products in the main category, a key attribute (size) was missing from the parameters; for two thirds of those it was only in the name, so a filter by parameter will not find it. | A customer cannot find the product with a parameter filter even though the shop sells it. |
| Catalogue structure (B) | Duplicate categories with the same purpose whose content diverges, and an empty category in the sitemap. | Split relevance and wasted crawling. |
| SEO (B) | About one in eight products shared its name with another, for about half the company name was inserted in the middle of the page title and about every seventh had a duplicate meta description. | The search engine tells pages apart less easily and the result is less appealing. |
| GEO and AI citations (B) | I found no structured data about stores and no reviews. | Fewer verifiable facts for Google and AI assistants to draw on. |
| Legal (B) | The terms and conditions named a different domain and referred to a repealed law. | An untrustworthy impression and legal uncertainty. |
On shop A, after trial batches, I corrected the category and availability mapping in the feeds, added new descriptions and names and verified the result via the API and on the live website. For shop B it was an audit from the public website, so the result was a list of findings.
Both shops were up and running and publicly available. The findings still went unnoticed because nobody independent measured them.
It is not an exception
My public measurements give a similar picture. They are samples, not representative surveys, but they show that technical and content care tends to be weaker than you would expect:
- In the measurement of 19 online shops from the Moravian-Silesian Region (25 October 2026), the search an AI assistant uses returned 105 results for 12 national "category e-shop" queries and none was one of the measured shops. Availability on the product was shown by 8 of 15 shops, delivery time by 5 of 15. Median mobile Lighthouse performance was 44 out of 100 and 12 of 19 scored below 50.
- In the measurement of 21 Czech online shops, 8 of 20 active shops, or 40%, had an Organization or Store type in structured data.
- How Czech online shops are set up for AI crawlers I measured in AI crawlers and 30 Czech online shops.
- And in the measurement of 598 Ostrava company websites, 93 of them (15.6%) were unavailable at the time of measurement.
12 questions you can ask your agency
A good agency will answer them quickly and without embarrassment. If an answer takes weeks or sounds like "that's technical", it is a reason to look closer. Google recommends similar questions itself: in its Search Central documentation it advises asking an SEO provider, among other things, for examples of their work, whether they follow Google Search Essentials, how they measure success and whether they share every change they make to your site.
| Question | Good answer | Warning sign |
|---|---|---|
| 1. When did you last check the feeds, and how many products are in them compared with the catalogue? | A specific date and two numbers (in the catalogue and in the feed). | "Feeds generate themselves." |
| 2. How many categories have export to feeds switched off? | A number and the reason it is off. | They cannot answer. |
| 3. Do I have access to Search Console and Merchant Center on my own account? | Yes, accounts belong to the owner's company. | The agency owns the accounts. |
| 4. What changes did you make in the last three months and where is the list? | A dated list of changes with reasons. | "We keep improving things." |
| 5. How do you measure where orders came from? | Measurement in GA4 or an equivalent, with described attribution. | Only an estimate or "from Google". |
| 6. How many products have a duplicate or empty description? | A number and a plan to fix it. | "Descriptions come from the goods supplier." |
| 7. What does the shop look like on mobile, and have you measured it? | Measurements (Lighthouse, Search Console) and a result. | Checked only on a computer. |
| 8. Does the cookie banner match consent? | Cookies are not stored without consent, verified by a test. | "The banner takes care of it." |
| 9. Does the shop appear in AI answers to questions in your field? | A manual test and Google and Bing reports, with no promises. | "We don't care about AI" or a guarantee of citations. |
| 10. Who owns the accounts, domains and data? | You do, with a written overview of access. | A vague answer. |
| 11. How do you measure success and how often will I get a report? | Agreed indicators that match the goal (revenue, not just traffic). | Only traffic and positions. |
| 12. Can I have the shop checked by someone else? | "Of course, we will provide the material." | Refusal or nervousness. |
No single answer proves anything. What matters is the overall picture and whether the agency can back its answers with evidence.
What you can check yourself in 15 minutes
None of this requires technical knowledge:
A 15-minute shop check
- Feed versus catalogue find out in the comparison site's administration how many products it received and compare that with the number in your shop
- Mobile open the shop on a phone, go through the menu to a subcategory and follow one product from price to basket
- Your own category search for your main category (for example "category e-shop" in Czech) and see whether you appear, or only directories and comparison sites
- Search Console in the Performance report, see which queries categories and products show up for, not just the homepage
- Account ownership check that you can get into Search Console and Merchant Center under your own account
A more general website check is in How to tell a bad website. For a quick automated test, use the free SEO audit.
Why an independent audit
Questions and a quick check will uncover the main things. To go deeper you need a second opinion from someone who does not manage the shop, because they are not grading their own work and have no reason to hide or exaggerate anything.
What an audit should contain: for each finding a description, evidence (a file, an address, a number), the impact and a priority. What such a report looks like and what to expect from it is in my article What an SEO audit reveals.
What to do with the result: show it to your current agency first. Often it will turn out to be things nobody had mapped, not bad intent. Change suppliers only if the problems remain and nobody is dealing with them. An audit is not an attack on the agency; it is the basis for a conversation.
E-shop audit for €280. I only need access to your Shoptet account. I go through the template, feeds, categories, product data and tracking, and you get a ranked list of what to tackle first. I deduct the audit price from follow-up work if you decide to order it within 60 days. Access stays yours. Shop outside Shoptet? Have a look at the SEO audit: an audit of a website starts at €400, and an e-shop is quoted by scope.
Frequently asked questions
How do I tell that an agency is neglecting my shop?
There is no single sure sign. Warning signs are answers without numbers and data, being unable to get into your own accounts, reports that cover only traffic and changes that nobody records. Certainty comes only from measuring: feeds, mobile, cookies and page titles.
Is an audit from another agency unfair to the current one?
No, as long as it is factual and based on evidence. The audit is a basis for a conversation. Show it to your current agency first and give them room to fix or explain. That is fair to both sides.
How much does an independent audit cost?
With me, an audit of a Shoptet shop costs €280 and I deduct it from follow-up work if you decide to order it within 60 days. An SEO audit of a website starts at €400; an e-shop outside Shoptet is quoted by scope. I quote a larger or more complex shop individually.
Do you need access for the audit?
For Shoptet, yes: an e-shop audit only needs access to your account. Access stays yours and you can remove it at any time. Some findings can be established from the public website, but feeds, category settings and the administration cannot be seen without access.
What if the audit finds nothing?
Then you have confirmation that the shop is in good shape and a certainty you did not have before. The result can also read "this is fine, here are two minor things". There is nothing wrong with that; it is useful information too.
Will you promise better positions or AI citations?
No. Nobody can reliably guarantee positions or citations in AI answers. An audit shows what prevents search engines and assistants from reading you well and orders the steps by impact.
Sources and links
- Do you need an SEO?, Google Search Central
- Online shops from the Moravian-Silesian Region, measured on 25 October 2026
- I measured 21 Czech online shops
- AI crawlers and 30 Czech online shops
- Ostrava company websites 2026
The examples of shops A and B are anonymised and given as ranges. Aggregated background data is available from me on request, without names and domains.