MAR & disclosure details
Does the announcement include the disclosure details readers need?
Publication details still to be added → classification, issuer and time included.
Method and sources
The detail behind the case. PCS estimates are preliminary, the AI test showed no improvement, and the website rating is a dated observation.
The working material becomes a finished announcement through the company’s additions, geological review and editorial work. Here is one part of that change.
Reviewer scores · 0 to 100
Preliminary internal reviewer scores. The scale has not yet been validated. The MAR score describes the publication text, not the percentage of legal compliance.
Does the announcement include the disclosure details readers need?
Publication details still to be added → classification, issuer and time included.
Do the figures reconcile?
An interval length does not reconcile → sample lengths and weighted intervals checked.
The change includes the company’s additions, geological review and Pinecap’s editorial work. A working draft and a published announcement serve different purposes.
A concrete sentence example
More drilling is needed to establish the extent of the mineralisation. That point becomes easier to find.
From the published announcement · translation
Further drilling is required to establish its width, length and dip.
The starting point is summarised. The published sentence is translated from Swedish.
Initial material
A calculation needs correction. Conditions are buried in the text. Publication details are still to be added.
Published announcement · 26 June
Sample lengths and intervals reconcile. The need for further drilling is stated directly. The announcement identifies its issuer and date.
News value
The main finding and its limitations are clearer.
Links to milestones
The announcement connects more clearly to prior work and next steps.
Room for misinterpretation
Down-hole length is distinguished from true width. Limitations are made explicit.
Decision support
Timing and dependencies are clearer. Further drilling has not been decided.
PCS weighs six areas: news value, numerical consistency, links to milestones, room for misinterpretation, decision support and MAR hygiene. Disclosure details include information needed for publication; a PCS score is not a percentage of MAR compliance.
The checks shown cover 35 sample lengths, table continuity and four length-weighted intervals in the public original. They verify arithmetic, not the geology or laboratory certificates. A working draft and a final announcement serve different purposes in the process.
The dimensions show the mean of two documented internal assessments of the Swedish material, made for this example. Half-points result from averaging reviewer scores, not precise measurement. The English view translates the same assessment. Colour follows the score continuously; it is not an approval threshold. The collaboration page also shows the historical internal aggregate assessment of 29 to 83. Inconsistencies in the method’s calibration references must be resolved before scores can be used as external evidence. .
Company facts need to be easy to find and quote in context, including when the reader uses an AI service.
Implemented on the website
Press releases identify the issuer and date. The glossary contains 12 structured definitions. A dedicated source guide points to the company’s original material.
Actual AI answer in the reading test
How is the mineral resource of 4.6 tonnes of gold classified?
“Stortjärnhobben’s mineral resource is stated as 4.6 tonnes of gold and classified as Inferred under PERC 2021.”
Answer citing the project page. The older material also produced this correct answer. Translated from the Swedish test. Question clarified editorially; the recorded answer is unchanged.
Six identical questions, the same model and settings, three runs per set of material. The actual local backup from before the AI work was compared with the corresponding pages and source guide saved on 4 September 2026.
| Material | Correct answers | Runs |
|---|---|---|
| Before · earlier website | 5/6 | 3 |
| After · 4 September | 5/6 | 3 |
This test did not establish an improvement in answer accuracy. The central geology answers were correct with both sets of material. The older material provided no answer about registered office and headquarters. The new source guide’s incorrect registered-office location was repeated in all three runs. The guide therefore needs to distinguish registered office in Stockholm from headquarters in Skellefteå.
The test covers three pages and the source guide, supplied directly to gpt-5.6-sol with the same high setting. No external searches were used. It does not measure historical ChatGPT visibility, indexing or answers across all AI services. The source material and actual answers have been retained for review.
Clearer browser instructions for encrypted connections, embedding and permitted content sources.
Security Headers
The follow-up shows all six tested settings in green. A concrete improvement, documented in the report dated 2 September.
The grade applies to these settings, not to the security of the entire website.
Same service, Security Headers, and the same address, lapplandguld.se. Before: 25 August, reproduced in ClearAgency’s report. After: 29 August, reproduced in Pinecap Audience, sent on 2 September. The exact report attachment has been verified.
Added settings: Strict-Transport-Security, Content-Security-Policy, X-Frame-Options and Permissions-Policy. A historical result, separate from the AI reading test.