How we decide what to publish

Our seven commitments

1. 1. We start from primary sources.

   Card fees, rewards, eligibility, and terms come from the bank's own MITC document, the official product page, or an RBI circular. Aggregator and comparison sites are never the source — they are at best a cross-check. If a fact cannot be traced to a primary source, we do not publish it.

2. 2. We separate fact from opinion.

   Numbers, dates, eligibility, fees, fine print — these are facts. Whether a card is "best" for someone is opinion, and opinions are framed against a stated user profile (e.g., "best for someone who spends ₹50,000/month on groceries"). We do not write open-ended "best of" claims without a user context.

3. 3. We will not invent benefits.

   If a card's benefit is unclear or undocumented, we say so. We will not assume "probably has lounge access" or "likely earns 4x rewards on dining" because the bank's marketing copy implies it. Ambiguity is reported as ambiguity.

4. 4. The score comes before the commission.

   CardTrail's scoring algorithm has no input for "do we have an affiliate link." Cards without affiliate relationships get the same scoring rigour as cards with them. A high-scoring card may have no affiliate; a low-scoring card may have one. The score does not move because of this.

5. 5. We disclose conflicts.

   Where a card on our site has an affiliate relationship, it is marked. Where the editor or anyone close to them holds the card personally, that is mentioned in the article. Where the bank has provided sample data or media access, that is stated.

6. 6. We update with the source.

   When a bank changes a fee, devalues a benefit, or amends terms, the article reflecting that fact is updated within seven days of the change being published in primary sources. Updated articles carry a "Updated" date that reflects the actual edit, not the original publication.

7. 7. We correct in public.

   If we publish something wrong, we issue a correction note inside the article. We do not silently edit. The correction states what we got wrong, what is now correct, and the date of correction.

What we will not publish

• ✗ Sponsored "best card" rankings. No bank or card network can pay to alter a ranking. We do not publish "top X cards" lists where the order is influenced by commercial relationships.
• ✗ Recycled marketing copy. If an article reads like the bank's brochure, it does not get published. Every claim is independently verified or removed.
• ✗ "Apply now" pressure tactics. Time-pressure language ("limited time offer," "apply before this rate ends") is reserved for cases where the bank itself has set a hard deadline, with the deadline disclosed.
• ✗ Speculation as fact. "This card is likely to add lounge access" is speculation. We label it as speculation or do not publish.
• ✗ Anonymous opinion pieces. Every article on CardTrail is signed by a named editor accountable for the content.

Our verification process

1. Source check. Every fact in the article is traced to a primary source URL. Where the URL is a PDF (MITC documents, RBI circulars), the PDF is archived in our internal record.
2. Date stamp. The date the bank's primary source was last verified is recorded. Articles citing a fee or benefit are stamped with the date that fee or benefit was confirmed live.
3. Cross-check on contested claims. Where two banks claim a comparable benefit (e.g., both claim "unlimited lounge access"), we read both MITC documents in full to surface the differences in fine print.
4. Editorial review. Every published article is reviewed by the editor for tone, claim integrity, and adherence to these standards before it goes live.
5. Post-publication audit. Articles older than 90 days are re-checked against current primary sources. If anything has changed, the article is updated and the "Updated" date refreshed.

Retraction and correction policy

If we publish a factual error — a wrong fee, a benefit that doesn't exist, a date that's incorrect — we follow this protocol:

1. The incorrect statement is corrected in the article body.
2. A visible correction note is added at the top of the article: what was wrong, what is now correct, and the date.
3. The article's "Updated" date and dateModified schema field are refreshed to the correction date.
4. If the error materially altered a recommendation (e.g., we recommended a card whose actual fee is higher than what we wrote), readers who reached the page through the email newsletter are notified in the next dispatch.
5. Severe errors (factually wrong scoring, misattribution of a benefit to the wrong card) result in the article being temporarily unpublished pending re-verification.

We do not silently edit factual claims. If you spot an error, write to the editor — contact details are at the bottom of every article.

Our policy on AI and automation

AI tools assist our research workflow — searching public databases, drafting structural outlines, and surfacing changes in MITC documents that humans might miss. They do not replace editorial judgement.

• · Every article is reviewed by a human editor before publication. Where AI assisted the draft, the editor verifies every factual claim against primary sources before signing off.
• · No article is published without a named editor. The byline reflects accountability — if a fact is wrong, the editor named on the byline owns the correction.
• · AI output is never published unedited. Drafts are restructured, fact-checked, and rewritten by the editor. Anything that reads like a generic LLM output has not passed our review.