Corrections Policy
Last updated: 8 September 2026
We correct factual errors as soon as we find them or as soon as a reader tells us about one. This page explains what counts as a correction, how we handle it, and how to report a problem. It also lists the corrections we have made.
What counts as a correction
A correction applies when something we published was factually wrong at the time it was checked, for example an incorrect price, an outdated feature, a misattributed source, or a regulatory citation that no longer matches the current guidance. It does not apply when a vendor simply changes their product after we checked it. That is normal drift, not an error, and is handled through our routine content refresh instead.
It also does not cover ordinary editing. Rewording a sentence, replacing a general link with a more specific one, or expanding a section with sourced detail are improvements, not corrections, and listing them here would make the record look fuller while telling you less.
How we handle a correction
When we confirm an error we update the affected page directly, record what changed and the date, and add it to the log below with the source that settled it. We do not remove or rewrite history to hide that an error existed. Corrections are additions to the record, not replacements of it.
Some pages carry their own visible changelog or a "last verified" date, and those are updated too. Neither is present on every page, so this log is the complete record rather than a summary of what those fields show.
Corrections made
33 corrections across 23 pages, newest first. Each entry names what was wrong, what the page says now, and the source we checked it against, so you can verify the correction rather than take our word for it.
8 September 2026
AI Change Management for Australian Businesses
What was wrong: The page pointed readers to a Fair Work Ombudsman page on modern-award consultation terms that no longer exists. The URL returned an error page, so a reader following it got nothing.
What it says now: It points to the Fair Work Ombudsman's current best-practice guide on consultation and cooperation in the workplace, which covers the same consultation obligations.
AI Tools for Australian Healthcare Practices
What was wrong: The page pointed readers to a TGA page on Software as a Medical Device that no longer exists.
What it says now: It points to the TGA's current guidance on clinical decision support system software regulation, which is the part of the TGA's material the surrounding paragraph is actually about.
AI Tools for Australian Healthcare Practices
What was wrong: An FAQ pointed readers to a short OAIC URL for APP 1 that no longer resolves.
What it says now: It points to the OAIC's current APP Guidelines chapter for APP 1, the same document at its present address.
AI Chargeback Evidence for Ecommerce Stores
What was wrong: A link to the European Commission's summary of GDPR data-processing principles no longer resolved.
What it says now: It links the Commission's current page on the legal grounds for processing personal data, which covers the same lawful-basis material.
2 September 2026
AI Governance in Canada: What PIPEDA, the OPC and Quebec Law 25 Actually Say
What was wrong: The page described Canada's Artificial Intelligence and Data Act (AIDA) as pending legislation, asking what it "would add if it passes". AIDA cannot pass: it died on the order paper when Bill C-27 lapsed at the January 2025 prorogation. The same page's own body already said so, so the article contradicted itself.
What it says now: It now refers to what AIDA would have added before it lapsed.
Source: https://www.parl.ca/legisinfo/en/bill/44-1/c-27
AI Governance in Canada: What PIPEDA, the OPC and Quebec Law 25 Actually Say
What was wrong: The page stated that under PIPEDA a given decision "counts as an automated decision". PIPEDA contains no such defined category. A right to an explanation was proposed in the CPPA, which died with Bill C-27, and exists in Quebec's Law 25 and GDPR Article 22, but not in current PIPEDA.
What it says now: It now says PIPEDA has no separate category of automated decision, while its accountability and openness principles still apply to how a decision is made and explained.
Source: https://laws-lois.justice.gc.ca/eng/acts/P-8.6/FullText.html
AI Governance by Region: International AI Regulation Compared
What was wrong: A comparison of privacy regimes said the GDPR, UK GDPR, PIPEDA/CPPA, the Australian Privacy Act and US state laws "all apply to their respective data subjects regardless of where the data is processed". Two errors: the CPPA is not law, and the Australian Privacy Act does not reach most businesses turning over $3 million or less.
What it says now: It now says each regime reaches data processed outside its own borders but each has its own coverage test, and names Australia's as the narrowest.
AI Governance by Region: International AI Regulation Compared
What was wrong: The maximum civil penalty under the Australian Privacy Act was given as "up to AUD 50M per breach". Section 13G sets it at the greatest of $50 million, three times the benefit obtained, or 30 per cent of adjusted turnover, so the figure understated the cap.
What it says now: It now states the greatest-of formula in full.
AI Governance by Region: International AI Regulation Compared
What was wrong: Australia's AI policy department was named DISER. It has been the Department of Industry, Science and Resources (DISR) since 2022.
What it says now: The department is named DISR.
Source: https://www.industry.gov.au/
AI Governance by Region: International AI Regulation Compared
What was wrong: The page said APP 8 applies "if you are Australian and using US-hosted AI tools". APP 8 is triggered by disclosure to an overseas recipient, not by where a tool is hosted, and it only applies to entities covered by the Privacy Act.
What it says now: It now says APP 8 applies where you are covered by the Privacy Act and actually disclose personal information to an overseas recipient, and that US hosting alone does not establish that.
AI Tools for Australian Healthcare Practices
What was wrong: The page presented consent and a substantially-similar-law belief as the two routes to complying with APP 8. That inverts the rule and its exceptions: APP 8.1 is a reasonable-steps obligation, and those are limited exceptions to it.
What it says now: It now states the reasonable-steps obligation first and describes the exceptions as exceptions.
AI Tools for Australian Healthcare Practices
What was wrong: The page cited an OAIC guidance URL that returned HTTP 404. A citation a reader cannot follow is not a citation.
What it says now: It links the current OAIC guidance on privacy and commercially available AI products, checked on 2 September 2026.
OAIC AI Guidance for Australian Business: What the Privacy Regulator Has Said
What was wrong: The page said businesses must take reasonable steps to ensure an overseas recipient protects information "under standards comparable to the APPs". APP 8.1 requires reasonable steps to ensure the recipient does not breach the APPs, which is a different and stricter test.
What it says now: It now states the obligation as written in APP 8.1.
AI Governance in the United Kingdom: Which Regulator Says What
What was wrong: ICO guidance was called "the authoritative reference for data protection obligations". ICO guidance is the regulator's non-binding interpretation; the obligations sit in the UK GDPR and the Data Protection Act 2018.
What it says now: It now says the binding obligations sit in the legislation and describes the ICO guidance as the regulator's interpretation and the practical starting point.
Source: https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/
AI Governance in the United Kingdom: Which Regulator Says What
What was wrong: The page attributed the UK automated decision-making Regulations to Article 22C UK GDPR and section 50C of the Data Protection Act 2018. Those provisions supply the definition; the enabling power is sections 124A and 124B.
What it says now: It now cites sections 124A and 124B as the enabling power and explains what Article 22C and section 50C actually do.
Source: https://www.legislation.gov.uk/uksi/2026/425/made
21 August 2026
How to Set Up Dext for a Bookkeeping Practice
What was wrong: The page advised reserving publishing rights to a smaller group of practice staff through role settings. Every practice user with access to a client account acts as an Admin inside it, so practice roles alone cannot achieve that separation.
What it says now: It now says Dext gives every practice user Admin-level access inside any client account they can reach, and advises limiting which clients each colleague can access instead.
Source: https://help.dext.com/en/articles/215320-roles-and-permissions-in-dext
TPB AI Guidance for Tax and BAS Agents
What was wrong: The page presented approved-use boundaries, access controls and escalation rules as what supervision of AI-assisted work requires. The TPB's final AI guidance does not prescribe those specifically for AI.
What it says now: It now says what the guidance actually requires, points to the TPB's broader supervision and quality-management material, and presents the specific controls as practices a firm may choose.
Source: https://www.tpb.gov.au/tpbgs-552026-use-artificial-intelligence-and-code-professional-conduct
Customer and Job-Site Data in AI Trade Tools
What was wrong: The page linked the original EU AI Act text. Regulation (EU) 2026/1744 amended the Act on 27 July 2026, so that link was no longer the current consolidated version.
What it says now: It links the current consolidated text and states the application dates it gives.
Source: https://eur-lex.europa.eu/eli/reg/2024/1689/2026-07-27/eng
How to Automate Bank-Reconciliation Review
What was wrong: The page said a small pilot "can usually be configured in a few focused sessions". No source supports that as a general estimate; vendor onboarding ranges from self-paced setup to XBert's four one-hour Fast Track sessions plus preparation.
What it says now: It now says the time and difficulty depend on the number of accounts, client files, integrations, rules and permissions, rather than offering a typical figure.
Source: https://support.xbert.io/en/articles/9334513-us-fast-track-program
How to Automate Client Document Collection
What was wrong: The page told readers to allow "roughly half a day" to design and configure the workflow. That universal estimate is not supportable against current setup guidance, which covers practice settings, users, permissions, client import and accounting connections.
What it says now: It now scopes the effort to client count, software, integrations and the complexity of the existing process.
Source: https://help.dext.com/en/articles/416727-getting-started-with-dext-for-accountants-and-bookkeepers
XBert Review for Accounting Practices
What was wrong: Cancellation advice told readers to confirm what happens to practice data when a subscription ends, without saying what XBert's terms provide. The terms do not promise post-termination export access, so a reader following the original advice could have lost data.
What it says now: It now says to export the Client Work Data Export for each client before cancelling, and states that access ends on termination.
Source: https://www.xbert.io/terms
19 August 2026
Best AI Phone Tools for Trades
What was wrong: The page implied ServiceM8's Phone Agent works from existing job and scheduling records. ServiceM8 states the agent does not have access to existing job information: it can take details and submit an enquiry, but it does not select a live schedule slot or promise a time.
What it says now: It now describes what the agent can and cannot do, quoting ServiceM8's own support documentation.
Best AI Phone Tools for Trades
What was wrong: Sameday was described as requiring a sales conversation for commercial terms. Sameday publishes plan pricing: Launch from US$449 a month for 500 minutes and Scale from US$789 a month for 1,000 minutes.
What it says now: It now says published plans have starting prices, while Enterprise and some custom integrations still require a sales conversation.
Source: https://sameday.ai/pricing
Best AI Bookkeeping Automation Tools
What was wrong: Botkeeper was characterised principally as outsourced or managed bookkeeping capacity. Botkeeper offers Infinite as a software-only product managed by the accounting firm, with dedicated human resources as optional add-ons.
What it says now: It now separates the software-only licences from the optional service tiers and says what each should be compared against.
Source: https://www.botkeeper.com/pricing
Best Field Service Scheduling Software
What was wrong: Pilot guidance assumed every shortlisted product offered an equivalent trial. simPRO says it does not provide a self-serve trial and AroFlo directs prospects to a guided demonstration, so a reader following the advice could not have run the pilot as described.
What it says now: It now names which products offer self-serve trials and their lengths, and which offer demonstrations instead.
Source: https://www.simprogroup.com/pricing
Best Month-End Close Tools for Bookkeeping Firms
What was wrong: Numeric was suggested for firms handling outsourced-finance engagements. Its standard public terms limit use to the licensee's internal business purposes and prohibit service-bureau use, so that suggestion cut against the vendor's own terms.
What it says now: It now frames Numeric as suited to an internal finance operation and states the service-bureau restriction.
Source: https://www.numeric.io/legal/terms
Best Receipt and Invoice Capture Tools
What was wrong: The page referred to "Dext Prepare". Dext combined Prepare and Precision into a single product now branded Dext.
What it says now: The product is named Dext throughout.
Source: https://dext.com/en/blog/single/the-next-phase-of-the-dext-evolution-is-here
Best Receipt and Invoice Capture Tools
What was wrong: The page said the GDPR "may be relevant to personal data connected with the European Economic Area". A connection between data and the EEA is not the territorial-scope test.
What it says now: It now follows European Commission guidance: processing through an EU establishment, or offering goods or services to, or monitoring, individuals in the EU.
Best AI Software Stack for a Small Accounting Firm
What was wrong: The page said the GDPR "may be relevant when personal data connected with people in the European Economic Area is processed", which is broader than Article 3's territorial-scope test. It also treated the EU AI Act as wholly prospective when parts are already in force.
What it says now: It now states the Article 3 test and links it, and describes the AI Act's provisions in force alongside its later transition dates.
Source: https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32016R0679
Using Client Financial Data in AI Tools
What was wrong: The GDPR was described as applying to practices "serving people in or connected with" the EEA. That is too vague to be the territorial-scope test.
What it says now: It now attributes the test to European Commission guidance: establishment in the EU or EEA, or offering goods or services to, or monitoring, people in the territory.
Best AI Tax Research Tools for US Firms
What was wrong: The page named Bloomberg's research feature "Bloomberg Tax AI Assistant". Bloomberg's current product material identifies it as Bloomberg Tax Answers; the earlier name belonged to a development feature released to selected customers in 2024.
What it says now: The product is named Bloomberg Tax Answers.
Source: https://pro.bloombergtax.com/insights/tax-automation/ai-for-tax-professionals/
ServiceM8 vs Tradify for Small Trade Businesses
What was wrong: A comparison row credited ServiceM8 with purchase orders. ServiceM8 has no built-in function for creating and sending supplier purchase orders and directs customers to third-party add-ons.
What it says now: The row now covers supplier invoices, material costs and job-cost allocation, and asks whether a third-party add-on can meet a purchase-order requirement.
Source: https://support.servicem8.com/questions/general/servicem8-feature-generate-send-purchase-orders
ServiceM8 Review for Small Trade Businesses
What was wrong: Housecall Pro was offered as an unqualified alternative to ServiceM8. Its terms state the service is intended only for users in the United States and Canada.
What it says now: It now presents Tradify and Jobber as the general comparisons and names Housecall Pro as an alternative for US and Canadian readers, with the territorial restriction stated.
Source: https://www.housecallpro.com/terms/
What this log does not cover
In September 2026 we reviewed the whole site and merged or retired a number of pages. 45 of the corrections we made in that period were to 26 pages that no longer exist on their own: their URLs now redirect to a broader page that never carried the claim. We have not listed them individually, because an entry pointing at a redirect is not something you can check. They are recorded in our internal project files.
This log begins with corrections made from August 2026, when we started recording a source URL and a reason for every claim we changed. Earlier edits were not documented to that standard, and we are not going to reconstruct them from memory and present the result as a record.
How to report a problem
If you believe something we have published is inaccurate, out of date, or misattributed, please tell us through the contact page with a link to the page and a description of the issue. We read every report and aim to review it promptly.
Related pages
See our methodology for how we research, source and score, and our independence and disclosure policy for how commercial relationships are kept separate from editorial conclusions.