Skip to content

Troubleshooting

The first place to look is the dashboard: if the group Why nothing is audited is there, it names the one gate that has to be fixed first. If it is not there, the engine is working — then the cause lies with a single listing or on the way into the shop.


Why nothing is audited

The engine stops at four gates, and it always names the first one. In this order:

Order Message Solution
1 The Content Engine is switched off Switch Enabled on, on the setup card. Existing findings stay, they only age
2 No active license Activate the module CONTENTENGINE in the hub or start a demo — action Activate in Hub, FactBox License. In a sandbox the module runs without a license
3 There are no marketplace listings yet Connect a marketplace or shop and list items on it. What is audited are listings, not items — a full item master without a listing yields not a single audit
4 The scheduled job queue entry is missing or on hold Run Create Job Queue Entry and check the entry is in status "Ready". Until then findings only change when someone runs the audit by hand

If the fourth gate is open and still nothing shows up: check the scope. By default only listings that are online are audited; Audit Offline Items takes the rest along.

The scheduled run stays silent when the license is missing

The job queue entry runs, finds the license gate closed and ends without an error message — it is not supposed to write an error into the queue every day. The entry then looks healthy, and still nothing happens. The dashboard names the real reason.


Symptoms and solutions

Symptom Likely cause Solution
Copilot is not ready No AI resource on file, or the capability is switched off Enter endpoint, deployment and key in the AI Setup and switch the capability Listing Content Draft on under Copilot & AI capabilities. Both are checked before a call starts
Copilot did not draft anything Budget exhausted, endpoint dead, or the answer was unreadable The attempt stays as a proposal in status Failed; the Error Text names the reason. On "not in the expected form" the deployment may not be a chat model
The batch stops mid-run Three failed drafts in a row By design: a fourth call against an exhausted budget or a dead endpoint would fail just the same. A message names the last error; proposals already created are kept
The draft could not be applied The license lapsed in the meantime, the proposal is already decided, or the listing no longer exists The draft waits in the content proposals and can be applied again from there. The message names the reason: Proposal … is … and can no longer be changed or The listing of proposal … no longer exists
A finding will not go away The listing has not been audited again since the correction Findings are only rewritten by an audit, not by saving. Use Audit SEO on the marketplace item card or Audit This Listing Again in the findings list — the daily run does it at the latest
A duplicate finding stays on the corrected listing The second listing still carries the text A duplicate always concerns at least two listings. While one of them holds the old text, the count is still greater than one — correct both and audit both
Description counts as missing although the shop shows text The text comes from a source the listing does not reach What is judged is what the channel serves: the listing's own text, otherwise marketing text, an external source, the item description. If the text lives only in the shop and nowhere in Business Central, the audit does not see it
Description too short although the page is long Visible characters are counted HTML tags do not count. A page of markup with little text is exactly the case the rule is about
Slug invalid although it works in the shop Capitals, umlauts, spaces or a doubled separator Shops turn such things into encoded characters or redirects; a search engine counts both against the page. Rules and examples: Findings
Texts do not arrive in the shop The connector has not synchronised yet Applying writes the hub fields and sets the update flag on the listing. The texts go out with the connector's next item update — not immediately. If that synchronisation does not run, check it on the connector in question
Texts in the shop, but not in the meta fields Connector version too old The SEO fields are channel-neutral in the hub; the connector has to know them to map them (Shopware 6 from connector 1.11 on to metaTitle, metaDescription, keywords, seoPathInfo)
Too many warnings after the first run The limits do not fit your assortment Every limit is a setting; 0 switches that check off. Change the setup and audit again

Frequently asked questions

Does applying overwrite my maintained texts? Only the fields the proposal actually carries something for. An empty proposed field leaves what is there standing — a cleared field means "keep what is there", not "delete it".

What happens when the engine is switched off? It audits nothing, drafts nothing and applies no proposal. The existing findings stay and age — they describe the state of the last run, not of today.

Why does applying depend on the license too? Because a draft costs the customer tokens, and because an applied text without the following re-audit would leave the listing with findings that no longer describe it.

Why is a listing skipped in the batch? It already has a proposal in status Proposed. No second draft while the first one is unread. Apply or reject it, and it counts again.

Can I use the module without AI? Yes. The audit is pure rule evaluation over hub tables, with no external service and no credentials. Only the Copilot drafts need Azure OpenAI.

Will findings and proposals become too many over time? Both tables are registered for retention policies — findings via Detected At, proposals via Created At. For the proposals the filter Status <> Proposed is recommended, so open proposals never disappear automatically.


Support

ALTENBRAND Datentechnik GmbH — support@altenbrand.de — www.merchantcentral.de