In the 2010s, the SEO authority signal that mattered most was backlinks. The page that earned the most links from credible sources won the rankings. Brands optimized accordingly. Whole industries were built around link acquisition.
The SEO authority signal that’s emerging now is catalog honesty.
This is the thesis we’ve been testing for the past year, with growing confidence. Agents reward brands whose product information is internally consistent, accurate against ground truth, and honest about limitations. Agents penalize brands whose product information is contradicted across the page, inflated against verifiable reality, or selectively positive. The signal isn’t subtle once you measure for it. It also compounds, because agents that get burned once start downweighting the source persistently.
This post is the long version of the argument. What “honest catalog” actually means, why agents check for it more aggressively than humans do, the specific dishonesty patterns that get penalized, and the practical work of building a catalog that earns the agent’s trust.
What “honest catalog” actually means
The phrase covers a few different specific properties that all need to be true.
Internal consistency. The product description, the PDP detail, the structured data (JSON-LD), the FAQ, the llms.txt, the /agent/query endpoints, and the policy pages all say the same thing about the product. Where one says “ships in 2 to 5 business days,” they all say “ships in 2 to 5 business days.” Where one says “rated -5°F to 25°F,” they all say “rated -5°F to 25°F.” The product’s structured data and its prose agree exactly.
Accuracy against ground truth. The claims the catalog makes are true. The jacket really is rated to -5°F. The reviews really do average 4.5 stars. The product really is made in Portugal. Stock status really matches the warehouse. The merchant who can defend every claim against an audit is a merchant whose catalog passes the agent’s accuracy check.
Honest framing of limitations. Products have weaknesses. Catalogs that name them explicitly (the jacket runs small, the lipstick discontinues in certain shades, the speaker requires a separate power adapter) get treated as trustworthy by the agent. Catalogs that hide weaknesses or bury them in fine print get treated as promotional, which is a low-trust signal.
Balanced review aggregation. AggregateRating that reflects the actual distribution, not a curated subset. Written reviews that include both positives and negatives, not just five-star testimonials. The catalog whose reviews surface real problems is a catalog the agent can confidently quote in a recommendation, because the agent’s downstream user isn’t going to discover something surprising after the purchase.
A catalog that has all four properties is rare. It’s also disproportionately rewarded.
Why agents check more aggressively than humans
Humans visit one site at a time. They have limited ability to cross-reference claims across sources in real time. They take the merchant’s word for most things and trust their gut on the rest.
Agents do all the work humans don’t. They cross-reference structured data against prose on the same page. They cross-reference one merchant’s claims against another merchant’s claims, against third-party reviews, against industry publications. They check whether the Offer block in JSON-LD agrees with the price visible on the page. They check whether the temperature rating in the structured data matches the temperature rating in the marketing copy. They flag inconsistencies, downweight the source, and remember the pattern.
This is a feature of agents, not a bug. Agents are built to give the buyer a confident answer. A confident answer requires sources the agent can quote without misleading the buyer. The agent’s check on catalog honesty is in service of the agent’s reliability to its own user. It will only get more aggressive as agents get more sophisticated.
The implication for merchants is direct. The honesty work that used to be optional (or that traditional CRO advised against because it might cost you a conversion) is now table stakes. Merchants who haven’t done the work are losing referrals to merchants who have.
The specific dishonesty patterns that get penalized
Across pilot tenant audits and the broader agent behavior research, we’ve catalogued the patterns agents consistently flag. The ones that matter most:
Cherry-picked comparisons. Marketing pages that compare your best feature against your competitor’s worst feature, or that select the metric where you win and exclude the ones where you don’t. Agents read these as low-trust and discount them. Worse, they often go fetch the missing comparison data from the competitor’s site and present it to the buyer alongside yours, which sometimes flips the recommendation toward the competitor.
Specificity asymmetry. Specific numerical claims for your product’s strengths and vague qualitative claims for your weaknesses. “Our battery lasts 18 hours of continuous playback” (specific) followed by “compatibility with most devices” (vague) is a pattern Gemini and Claude both flag. The agent reads “most devices” as a signal that the answer is selective.
Review distribution gaming. AggregateRating that doesn’t match the visible review distribution. Five-star reviews from accounts with no other activity. Star ratings that have been climbing suspiciously fast. Agents that compare your AggregateRating against third-party sources (industry blogs, marketplace listings, social mention sentiment) and find a gap will downweight you.
Competitor strawmanning. Subtly misrepresenting a competitor’s capability, often by referencing an older version of their product or an out-of-context limitation. The agent that fetches the competitor’s current spec sheet and sees the misrepresentation flags you, not the competitor.
Recency deception. Marketing copy that implies a competitive advantage that was true twelve months ago but isn’t true now. “The only product on the market that supports X” is true until it isn’t. Agents fact-check claims against current data. Stale superlatives are a fast way to lose trust.
JSON-LD scope parity. The JSON-LD covers a subset of what the prose covers, or vice versa. The structured data lists strengths but not limitations. The prose mentions a feature that isn’t in the structured spec. Agents that weight JSON-LD heavily (Gemini does this aggressively) get a different picture from the structured data than from the page text, and the inconsistency is the trust hit.
Each of these patterns is something a traditional marketing team has been doing for a decade as standard practice. Each of them is now a measurable disadvantage on the agent surface.
The compounding effect
Honesty isn’t a single signal. It’s a property of the entire information surface a merchant exposes. The compounding works in two directions.
In the positive direction: a merchant who’s clean across all of the honesty dimensions gets a confidence boost from the agent. The boost shows up as higher inclusion in shortlists, higher placement in recommendations, more frequent citation. Each interaction reinforces the signal. The agent learns that this merchant’s catalog can be trusted, and the trust gets remembered through the platform’s persistent memory features.
In the negative direction: a merchant who’s caught in any single dishonesty pattern gets a confidence hit. The hit isn’t just on the specific page or claim that was inaccurate. It propagates to the merchant’s entire catalog because the agent’s working model is “this source is unreliable, downweight everything from this source.” The merchant whose AggregateRating doesn’t match third-party data has all their products downweighted, not just the one with the inflated rating.
This compounding is part of why catalog honesty looks like the authority signal of the era. The brands that build the right catalog hygiene early compound favorably across years. The brands that don’t compound unfavorably.
The practical work
Building a catalog that earns agent trust isn’t theoretical. It’s a specific set of practices that hold up under audit.
Centralize your product spec source of truth. Every claim about a product (price, availability, temperature rating, dimensions, materials, weight, warranty, return policy) should live in a single canonical store. Every page (PDP, category, JSON-LD, llms.txt, FAQ, /agent/query endpoints) should read from that store. The cost is upfront engineering work. The benefit is that internal consistency is automatic instead of constantly drifting.
Audit your AggregateRating against the underlying reviews. Pull the actual review data. Compute the actual distribution. Make the JSON-LD match. If you’ve been inflating, the immediate hit looks like a worse-looking score on your PDP. The downstream benefit is higher agent trust, which translates to more referrals, which translates to more reviews from buyers who arrived with accurate expectations.
Name limitations directly in your product descriptions. “Runs slightly small.” “Requires assembly.” “Ships from Vermont, please allow extra time for West Coast orders.” The merchants who do this lose nothing in human-facing conversion (because buyers were going to find out anyway) and gain meaningful ground in agent-facing trust.
Run a quarterly honesty audit. Sample 20 of your top products. For each, check internal consistency across the page, JSON-LD, llms.txt, and FAQ. Check the claims against ground truth (actual specs, actual reviews, actual policy text). Check competitor mentions for accuracy. Document the gaps. Fix them. Re-audit in 90 days.
Wire the maintenance into your existing ops. The honesty work is only useful if it stays current. Build the cadence into your existing product launch flow, your existing review-management flow, your existing policy-update flow. The brands that treat honesty as a one-time deliverable see it decay over twelve months. The brands that treat it as a maintenance discipline compound on it.
What we offer
We built the Inception Honesty audit because the work above is concrete enough to systematize but tedious enough that most merchants don’t do it consistently. The audit runs across your structured data, your prose, your reviews, and your policy disclosure, flags the specific patterns above, and gives you a prioritized fix list. If you want to see what it surfaces on your catalog, the run is free at inceptionagents.com/audit.
You can do the work without us. The merchants who start with an internal audit and bring tooling in later still come out ahead of the ones who don’t audit at all. Either path beats the path most merchants are on, which is shipping the catalog they shipped a year ago and hoping the new agent-mediated surface treats it kindly. It won’t.
Catalog honesty is the new authority signal. The merchants who recognize that and build the discipline now are going to compound on it for years. The merchants who don’t will discover the cost the same way most things get discovered in commerce: when the numbers force the conversation.
Worth being ahead of.