How Can I Help

What Google's How Can I Help Actually Does

It is a conversational AI shopping surface from Google. When someone types a query like "what are good running shoes for flat feet under $100", the system can surface your product listings directly in the response instead of sending them to a traditional search results page. It pulls from your Merchant Center feed, matches inventory against natural language queries, and renders the products in a structured card format. The goal is to capture shoppers earlier in the decision funnel before they ever click through to a competitor or Amazon. There is no separate dashboard to log into. Everything is managed inside your Google Merchant Center account and tied to your Product Shopping campaigns. The first thing you need is a fully approved merchant center account with a complete product feed. If you have disapproved items or missing required attributes, the system will silently filter those products out of the How Can I Help results, and you will have no idea why your traffic dropped until you dig into the feed diagnostics page. After your feed is clean, you need Shopping or Performance Max campaigns running with the appropriate product data. Google automatically includes eligible products from active campaigns in the conversational shopping experience. There is no checkbox to opt in. You cannot opt out either, which matters more than people realize.

I learned this the hard way last year. A client had a feed with about 4,000 SKUs, and they noticed their Performance Max campaign impressions fell by roughly sixty percent overnight. They panicked and assumed the account was suspended. It turned out Google had started surfacing a large chunk of their products in How Can I Help responses, which meant the same inventory was getting impressions in the AI conversation layer without triggering the normal cost-per-click charges their reporting tracked. Their campaign spend looked flat because clicks were being routed to the free AI result instead. The fix was simply realizing that the traffic shift was real and adjusting their attribution window to account for the delayed conversion signal that comes through after a How Can I Help interaction.

Feed Quality Is the Real Bottleneck

The system relies heavily on product title structure, descriptions, and Google Product Category codes. Generic titles like "Blue Shirt Men's Size M" will get buried. Descriptions with keyword stuffing also hurt because the AI parses semantic meaning, not keyword density. I recommend cleaning up your titles to include the actual product use case. A title like "Men's Running Shoe with Arch Support for Flat Feet - Lightweight Cushioned Sneaker" performs noticeably better because it matches the way people actually phrase their search queries. Google Product Category is another area where most people fail. You do not need to guess perfectly, but using overly broad categories like just Apparel & Accessories instead of drilling down to Sports & Outdoors > Exercise & Fitness > Running reduces your visibility in the AI matching process significantly. The difference usually shows up within a week or two of the next feed refresh.

Common Pitfalls That Cost Money

Stock accuracy. If your feed reports a product as in stock but it is actually out of stock at the retailer level, customers who find your product through the AI interface will land on a dead page. Google penalizes this quietly by demoting your listings across all surfaces, not just the AI one. You need automated inventory syncs, ideally updating every few hours at most. Price competitiveness. The AI shopping responses tend to favor products with competitive pricing relative to similar items in the same category. This does not mean you have to be the cheapest. It means if five similar running shoes are showing and yours is twenty percent above the median, it gets pushed down the result set. I have seen clients adjust their pricing strategy just for this surface because the impression share loss was costing them more than the margin reduction. Brand gating. If you sell alongside your competitors' branded products and you do not have a brand attribute set correctly, your unbranded or incorrectly branded items get filtered out of premium placement. This is one of those counter-intuitive things nobody tells you about. Setting your brand field even for generic or store-brand items makes a measurable difference in how the AI surfaces your listings.

What It Cannot Do

How Can I Help does not replace your search campaigns. It does not give you granular keyword targeting. You cannot set bids on specific product queries that route to the AI interface. You also cannot see individual impression counts for How Can I Help specifically in your reports. The metrics are merged into your overall campaign data, which is why the attribution problem I mentioned earlier is so frustrating. The system also fails completely for certain product categories. Hazardous materials, age-restricted items, and products that require in-person verification are excluded. If your category sits in a gray area, your items may simply never appear regardless of how good your feed is. You will not get a rejection notice either. They just do not show up. If you are selling physical products with a compliant feed and you already run Shopping campaigns, there is nothing to download or configure for this. It is automatic. Focus your effort on feed hygiene and understanding that some of your traffic is now operating outside the standard click-based attribution model.