Google's Merchant Advisor Probably Can't Replace Your Judgment Yet

How much of your feed optimization would you hand to a vendor's own AI? Google is testing a feature called Merchant Advisor inside Merchant Center. It surfaces recommendations on product listings, titles, and attribute completeness. On paper, it sounds useful. In practice, you're asking the platform that sells you ads to also tell you how to structure the inputs those ads depend on. That's not inherently corrupt, but it is a conflict of inference worth naming.

What Merchant Advisor Actually Does

Based on early reports from Search Engine Land, the tool appears to function as a diagnostic layer. It flags feed issues. It suggests attribute improvements. It probably uses the same quality signals Google already applies when scoring your product listings for Shopping placements. None of this is new logic. Google has offered feed diagnostics for years through automated rules and disapproval notices. The difference now is packaging. Rather than a list of errors, you get something resembling a conversational interface with recommendations ranked by estimated impact.

The risk isn't that the advice is wrong. It's that the advice is narrowly optimized. Google's model is trained on what performs inside Google's ecosystem. It has no visibility into your margin structure, your inventory constraints, or your brand positioning on Amazon, TikTok Shop, or DTC. A recommendation to add a size attribute to improve impressions could be correct at the feed level and irrelevant at the P&L level. Operators who treat these outputs as hypotheses rather than directives will extract the most value.

The Query Report Problem Makes This Harder

Separately, Google acknowledged that Search Query Reports may not always show the actual searches users typed. This matters more than it sounds. Query reports are how commerce teams decide which keywords to bid on, which negatives to add, and which product titles to adjust. If the data feeding those decisions is an approximation rather than a transcript, your optimization loop has a calibration gap. You're tuning a system with noisy inputs.

Now layer Merchant Advisor on top. You have an AI tool making recommendations based on signals you can't fully audit, referencing query data that may not reflect real user behavior. That's two layers of inference between your budget and your customer's intent. Not fatal. But worth accounting for.

The Larger Context: Search Is Shrinking as a Channel

Condé Nast recently projected that search will become a single-digit percentage of its total traffic. Publishers and commerce brands are not the same, but the direction is shared. AI Overviews compress click-through. Chatbot-driven discovery routes users around traditional SERPs. Meanwhile, MIT Technology Review reported that AI chatbots are surfacing real phone numbers and personal data in responses, which raises questions about what information these models treat as public and how reliably they handle commercial queries. If an AI overview can hallucinate a phone number, it can probably hallucinate a product spec.

This convergence creates an environment where your dependency on Google's ecosystem is simultaneously increasing in complexity and decreasing in reliability. That's not a reason to leave. It's a reason to diversify your eval criteria.

The Operator's Move

Use Merchant Advisor. Don't obey it. The right decision for commerce operators is to integrate this tool into a broader audit workflow rather than treating it as an autonomous optimizer. Specifically, pull its recommendations weekly, cross-reference them against your own margin data, and test changes in controlled cohorts before rolling them across your full catalog. If Merchant Advisor says your titles need restructuring, run an A/B on 50 SKUs first. Measure incremental ROAS against your actual cost-to-serve, not Google's projected impression lift.

Build a simple scorecard. Three columns: what Google recommends, what your data supports, and what you actually implemented. Over 90 days, you'll have a calibrated view of how often the advisor's suggestions align with profitable outcomes. That scorecard becomes your moat. Most competitors will accept the defaults. You won't.

Second, pressure-test your query data. Export your Search Query Reports monthly and compare them against your own site-search logs, customer service transcripts, and social listening data. Where the gaps are widest, you're probably misallocating budget. This is manual work. It's also the kind of work that separates top-decile operators from median ones.

Third, reduce single-channel concentration. If search traffic is trending toward single-digit share for major publishers, your brand's search dependency deserves a hard look. Map your customer acquisition by channel quarterly. Any channel above 40% of new customer volume is a vulnerability, not a strength.

Three Questions to Pressure-Test

1. When was the last time you rejected a platform-generated recommendation based on your own margin data, and what did that decision cost or save you? 2. If your Search Query Reports are roughly 15% inaccurate, how large is the budget exposure you can't currently attribute? 3. What percentage of your Q3 acquisition plan depends on a single channel whose underlying data you cannot independently verify? One uncertainty I'll name: we don't yet know how Merchant Advisor weights its recommendations or what training data it draws from. If Google publishes eval benchmarks showing advisor-driven changes outperform merchant-driven changes across controlled cohorts with margin-adjusted metrics, I'd update my skepticism. Until then, trust the tool the way you'd trust any vendor pitch. Verify first.