How we measure what leading AI platforms recommend.

Credit Consensus Index measures AI recommendation consensus. It does not attempt to prove that artificial intelligence has identified the objectively best financial company or product. If multiple independent AI platforms recommend the same company for the same consumer situation, that agreement is worth documenting. A seven-platform study in which six platforms recommend the same company demonstrates stronger cross-platform consensus than a company recommended by only one platform.

Consumer-credit recommendations depend on context. A person with no credit history, a 580 credit score, a recent bankruptcy, multiple collection accounts, substantial credit-card debt, or inaccurate credit-report information may need completely different products or services. We therefore study narrower questions such as credit repair for consumers with multiple collection accounts, consolidation loans for borrowers with scores between 620 and 659, or credit-building products after bankruptcy. The variable tested should be significant enough to reasonably change the recommendation.

Our core categories include Credit Repair; Debt Relief; Personal & Debt Consolidation Loans; Credit Building & Rebuilding; and Credit Monitoring & Scores. Research may cover credit reports, collections, debt settlement, hardship, loan qualification, secured cards, credit-builder loans, rent reporting, FICO and VantageScore, monitoring, identity-related tools, and related consumer-credit services. Commercial intent is not hidden: we study realistic questions consumers ask when they are close to choosing a company or product.

Before collecting responses, we define the consumer situation, category, primary decision, material constraints, constant variables, and the variable that differentiates the study. A question should create a meaningful consumer scenario without being so restrictive that only one predetermined company can qualify.

Each platform should receive materially the same research question. We do not intentionally make one prompt more favorable to a company, insert affiliate partners unless the study explicitly concerns them, tell platforms who should win, or ask different platforms incomparable questions. Prompts are specific, neutral, commercially realistic, comparable, and open to competition. In open recommendation studies, we do not seed a predetermined list of companies.

The standardized question is submitted independently across the platforms included in the study. Our standard studies generally analyze seven independent consumer-facing AI answer and recommendation platforms, with the exact set identified on the study page and in Platforms We Analyze. Multiple platforms differ in models, training data, retrieval, browsing, source selection, ranking logic, safety systems, and recommendation behavior, so one system should not automatically represent the broader AI environment.

Whenever technically practical, queries are conducted in new or controlled sessions to reduce contamination from prior conversations, previously mentioned brands, personalized instructions, steering follow-ups, and unrelated account context. AI recommendation studies are snapshots. We record the research date, platform, query, response, recommended companies or products, position, reasoning, and relevant cited sources. Publication date is not necessarily the date AI research was collected.

Raw AI responses form the foundation of the dataset. We analyze them rather than asking another AI system to summarize everything into one answer. We may record the company or product, platform, list position, reason, consumer-fit reasoning, advantages, disadvantages, warnings, eligibility considerations, pricing and feature claims, and citations.

A company generally counts when the platform presents it as a meaningful option for the studied situation, using language such as recommended, best for, good option, strong choice, worth considering, suitable for, top pick, or one of the better options. An incidental mention, comparison, or example of something to avoid does not automatically qualify. Recommendations can be conditional; we may count them while preserving the qualification in the dataset.

AI systems may use parent-company names, product names, abbreviations, outdated names, alternate spellings, or names of renamed and discontinued products. Before calculating consensus, we standardize names, distinguish parent companies from products, separate products from one provider, resolve alternate spellings, and separate similarly named companies. We avoid combining recommendations that represent materially different products.

After normalization, we calculate how many platforms recommended each qualifying company or product. With seven platforms, 7 of 7 equals 100%, 6 of 7 is approximately 86%, 5 of 7 approximately 71%, 4 of 7 approximately 57%, 3 of 7 approximately 43%, 2 of 7 approximately 29%, and 1 of 7 approximately 14%. Coverage is generally the clearest and most important consensus signal. A company normally needs recommendations from at least two independent platforms for a full situation-specific fit review.

Coverage shows how broadly a company appeared; position shows how prominently it appeared. Two companies may each appear on six platforms, while one appears mostly in positions one through three and the other in positions three through seven. Our general hierarchy is recommendation coverage, recommendation position, situation-specific support, and material qualifications or limitations. Complete ranking rules appear on How We Rank.

Consensus can also exist in product characteristics. Platforms may independently emphasize three-bureau reporting, no annual fee, a reasonable deposit, an unsecured graduation path, low costs, minimum qualifying debt, program fees, duration, state availability, or credit risks. These recurring themes help explain the recommendation landscape.

Companies meeting the threshold may receive a deeper fit review. This is not a generic company review; it asks why AI systems recommended the company for this consumer and whether available factual evidence supports that reasoning. Depending on category, we may verify pricing, fees, disputes, creditor interventions, cancellation, state availability, debt requirements, program duration, APR, loan terms, origination fees, prequalification, inquiries, deposits, credit reporting, graduation, score type, bureaus, monitoring, alerts, subscription cost, and cancellation terms.

Whenever practical, material financial claims are verified using current provider information. Other authoritative sources may include the CFPB, FTC, Federal Reserve, credit bureaus, FICO, state regulators, court or government records, and statutory or regulatory materials. AI citations are useful research data, but they are not automatic proof that a claim is correct.

AI systems can hallucinate, cite outdated prices, confuse products, describe discontinued features, misstate eligibility, misunderstand regulations, attribute features to the wrong company, or present old information as current. We distinguish what the platform said from what verification found. Accurately recording an incorrect AI response is not itself a Credit Consensus Index error; presenting that response as verified fact would be.

Consumer-credit research is reviewed by Mark B. Huntley, J.D. Mark co-founded CreditKnocks.com and authored hundreds of articles about consumer credit, credit repair, scores, reports, cards, loans, collections, debt, and credit-building before CreditKnocks was sold to FinMasters. He has personally experienced credit rebuilding and holds a Juris Doctor degree. He previously practiced law, is not currently a practicing attorney, and does not provide legal advice through Credit Consensus Index.

Human review may check the study question, prompt neutrality, practical relevance, product facts, outdated claims, credit concepts, consumer risks, conclusions, AI opinions, disclosures, and context. It does not give us permission to replace AI results with a preferred result. An affiliate partner does not receive extra recommendations, and a company we cannot monetize remains eligible to rank first.

Credit Consensus Index may earn money from affiliate relationships, lead generation, advertising, sponsorships, data licensing, and other commercial relationships. These relationships do not determine which companies appear in open AI recommendation studies. Our sequence is: research the market, calculate results, publish findings, then monetize eligible traffic where appropriate. A company does not need to be an affiliate partner to rank, qualify, receive a fit review, or rank highly.

The most specific ranking is an individual study for a defined consumer situation. Studies may also be aggregated by category using the number of qualifying studies, total recommendations, coverage, average position, breadth of situations, and changes across cycles. Overall leaders can reveal broad recommendation support, but no overall leaderboard means one company is right for every consumer.

AI recommendations change. Studies may be rerun using the same or materially equivalent scenario, producing new or lost recommendations, higher or lower consensus, different positions, new entrants, different reasoning, or different cited sources. Where practical, we preserve previous and current rankings, recommendation gain or loss, coverage change, research date, study version, and historical consensus.

A company does not need to change for its AI consensus to change. Platforms release models, change search partners and retrieval, update ranking logic, access newer content, alter source weighting, and change policies. Generative AI is not perfectly deterministic, and personalization may reflect account history, location, subscription, settings, search history, preferences, or experiments. Free and paid users, logged-in and logged-out users, countries, and product experiments may also receive different functionality.

Credit Consensus Index measures AI recommendations. It does not directly measure customer satisfaction, complaint rates, approval probability for an individual, long-term performance, regulatory compliance, solvency, every fee or restriction, whether a company is objectively best, or whether a consumer should purchase a product. Strong recommendation frequency is not proof of quality; platforms may share sources, favor visible brands, reproduce misconceptions, or underrepresent newer companies.

We aim to disclose the consumer scenario, research date, platforms, number of platforms, qualifying recommendations, coverage, ranking method, human reviewer, material limitations, and commercial disclosures. If we identify an error in analysis, transcription, calculation, or verification, we may correct it while distinguishing our error from an AI platform’s original error. Methodology changes should be documented, especially when historical datasets are not directly comparable. See Corrections & Updates.

Suppose seven platforms are asked for the best secured cards for someone rebuilding damaged credit after being denied unsecured cards. Product A is recommended by 7 of 7 platforms, Product B by 6 of 7, Product C by 5 of 7, Product D by 3 of 7, Product E by 2 of 7, and Product F by 1 of 7. Products A through E meet the two-platform threshold. A may be the clear consensus leader while B may be more practical if it requires a smaller deposit. Both observations can be true; consensus measurement and consumer-fit analysis are separate.

When a company is the #1 Credit Consensus choice, it means that, based on the platforms and research cycle in that study, it demonstrated the strongest cross-platform recommendation consensus under our methodology. It does not mean every consumer should choose it. Credit Consensus Index exists because AI systems increasingly influence financial decisions. Our goal is a transparent, repeatable record of what platforms recommend, where they agree or disagree, why they reach those conclusions, and whether material facts withstand human review.