Gate 2: Contract coverage. All must_cover_questions appear as H2 or H3. They do. No missing.
Gate 3: Claim = source. Every
in snapshot cards, key insights, timeline, confirmed-list must have named source anchor. Check the snapshot cards: each card has bullets with sources? The first card: bullets have CRA links. Second card: bullets have Western University link? Actually “2025 and 2026 YMPE figures are projections until officially confirmed later in the year” has a link to Western U. Third card: “2024 is the final enhancement phase…” has Western U link. Fourth card: “2025 YMPE projected at $71,300” has Western U link. All have sources. The timeline list items: each has a source link. The confirmed facts section: each has source link. So okay.
Gate 4: Fact lock vs verified_facts. Verified facts array is empty in input. So nothing to replace.
Gate 5: FAQ dedup. Need to compare FAQ items to H2/H3 text. Check for >85% lexical overlap. For example, FAQ “What is the difference between CPP and CPP2?” – this is not an H2/H3. H2s are different. So keep all. But need to ensure at least 5 after dedup. Currently 8 question items. No overlap issues.
Gate 6: Link hygiene + diversity.
– Internal links: There are two internal links in the “Related reading” FAQ item: “How to Invest in Stocks…” and “CIBC Investor’s Edge Login…”. Both are internal. Need to strip target and rel. Currently they have target=”_blank” rel=”noopener nofollow”. Remove those attributes.
– External links: Ensure all external have target=”_blank” rel=”noopener nofollow”. They do. Also ensure no duplicate external URLs. Check: many links to same CRA pages. Each URL used at most once? Actually there are multiple links to same URL. For example, https://www.canada.ca/en/revenue-agency/services/tax/businesses/topics/payroll/cpp-contribution-rates-maximums-exemptions.html appears many times. Gate says “Each external URL at most once.” That means each unique external URL should appear only once. That’s a problem. We need to deduplicate. For each unique URL, keep only the first occurrence, and for subsequent occurrences, either remove the link or link to a different relevant page? But contract says “Each external URL at most once.” So we must ensure no duplicate external URLs. However, the article repeatedly cites the same CRA page. We need to reduce to one link per URL. For subsequent citations, we can remove the hyperlink but keep the text? But gate says “each external URL at most once” and also “max 3 per domain”. So we need to manage. Let’s list all external URLs:
1. https://www.canada.ca/en/revenue-agency/services/tax/businesses/topics/payroll/cpp-contribution-rates-maximums-exemptions.html (CRA CPP) – appears many times.
2. https://www.canada.ca/en/revenue-agency/services/tax/businesses/topics/payroll/calculating-deductions/making-deductions/second-additional-cpp-contribution-rates-maximums.html (CRA CPP2) – appears multiple times.
3. http://www.uwo.ca/hr/pay/cpp.html (Western University) – appears multiple times.
4. https://www.canada.ca/en/services/benefits/ei.html (Service Canada) – appears multiple times.
All are from same domains: canada.ca (CRA and Service Canada both under canada.ca? Service Canada is also canada.ca) and uwo.ca. So max 3 per domain: canada.ca URLs count as same domain? Yes, canada.ca. There are 3 different paths under canada.ca: one for CPP rates, one for CPP2, one for EI. That’s 3, okay. But each path is used multiple times. We must ensure each URL appears at most once. So for each unique URL, we need to keep only the first occurrence and remove subsequent links (or replace with in-text citation). But the article is heavily reliant on these sources. We can remove the hyperlink from repeated citations but keep the source name as plain text? However, the gate says “Each external URL at most once.” It doesn’t say we can’t have the text without link. We can remove the anchor for duplicates. Also the internal links need target/rel stripped.
Let’s do a thorough link audit. We’ll process the HTML and for each external anchor, check if URL has been seen before. If yes, remove the anchor and keep the text (the source name) as plain text. Also ensure max 3 per domain. Since we have only 3 unique paths under canada.ca, that’s exactly 3. But we have also uwo.ca (one path). So total 4 unique URLs. That’s fine. But duplicates will be removed.
Also need to ensure that for the internal links in Related reading, we strip target and rel.
Gate 7: JSON-LD. Need to have exactly two
← Back to Blog