84% of Business Owners Now Use AI. Here's a 2-Minute Feed Check So AI Shoppers Can Find What You Sell

Robauto Team · Published · Last updated

Bank of America's Nov 17 report: 84% of business owners use AI and 68% of AI users expect it to lift sales. Here's a free copy-paste script that catches the feed errors that drop products from AI shopping answers.

Good news for Main Street landed this morning. In the 2026 Bank of America Business Owner Report, released November 17, 84% of small and mid-sized business owners say they use AI, and 94% of those users say it has helped their business in at least one way over the past year. Among owners using or planning to use AI, 68% expect it to increase sales and revenue in the year ahead.

The single most common use? Marketing and advertising, at 52%. That makes sense: getting found is where growth starts. And in 2026, getting found increasingly means getting found by AI shopping assistants that read your product data directly. So this post pairs the good news with a practical tool you can run today: a tiny, copy-paste script that catches the small product-feed mistakes that quietly drop products from AI shopping answers.

The Numbers: Small Businesses Are Leaning Into AI, and It's Paying Off

Two fresh surveys point the same direction. Bank of America's report (Ipsos survey of 849 small and 250 mid-sized U.S. business owners, fielded in August) and the OnDeck and Ocrolus Small Business Cash Flow Trend Report (860 small businesses, released November 3) both show AI moving from experiment to everyday tool.

FindingNumberSource
Business owners using AI in some capacity84%Bank of America, Nov 17
AI users saying it helped in at least one way94%Bank of America, Nov 17
Expect AI to increase sales and revenue next year68%Bank of America, Nov 17
Top AI use: marketing and advertising52%Bank of America, Nov 17
Owners planning at least one digital investment (AI tools lead at 50%)89%Bank of America, Nov 17
Small businesses using AI61%OnDeck/Ocrolus, Nov 3
AI users reporting a measurable positive impact90%OnDeck/Ocrolus, Nov 3
Owners expecting moderate to significant growth93%OnDeck/Ocrolus, Nov 3

Bank of America also found that owners see AI and hiring growing together: among those using or planning to use AI, 60% expect it to shift employees into new or redesigned roles and 52% expect to hire staff to manage AI tools, versus 33% who think it will reduce headcount. Overall, 39% plan to hire in the next 12 months. That's a constructive picture: AI as a growth tool, not just a cost cutter.

Where Growth Leaks: One Underscore in Your Product Feed

Here's the practical part. AI shopping assistants read product feeds, and each one has its own rules. A November 15 write-up from Logit Solutions AI spotted a perfect example: Google Merchant Center spells a pre-order item preorder, while the OpenAI product feed spells it pre_order. Same product, one underscore, and one channel can drop the row.

We checked both specs ourselves. The OpenAI product feed spec accepts only in_stock, out_of_stock, pre_order, backorder, or unknown, and states plainly: "Omitted, empty, or unrecognized values reject the row." Google's availability attribute lists in_stock, out_of_stock, preorder, and backorder.

RuleOpenAI product feedGoogle Merchant CenterSource
Pre-order spellingpre_orderpreorderOpenAI, Google
Unknown stockunknown allowedNot in supported listOpenAI, Google
Country fieldstarget_countries and store_country requiredNot part of this attributeOpenAI
GTIN format8, 12, 13, or 14 digits, valid check digit, no spaces or dashesChecked by the open-source tool belowOpenAI, shelfcheck

Copy-Paste: A 27-Line Feed Check (No Installs)

Save this as feedcheck.py. It uses only Python's standard library. Point it at a CSV export with availability, gtin, target_countries, and store_country columns.

import csv, sys

OPENAI = {"in_stock", "out_of_stock", "pre_order", "backorder", "unknown"}
GOOGLE = {"in_stock", "out_of_stock", "preorder", "backorder"}
TO_OPENAI = {"preorder": "pre_order"}
TO_GOOGLE = {"pre_order": "preorder"}

def gtin_ok(g):
    if not g.isdigit() or len(g) not in (8, 12, 13, 14):
        return False
    body, check = g[:-1], int(g[-1])
    total = sum(int(d) * (3 if i % 2 == 0 else 1)
                for i, d in enumerate(reversed(body)))
    return (10 - total % 10) % 10 == check

for n, row in enumerate(csv.DictReader(open(sys.argv[1])), start=2):
    a = row.get("availability", "").strip().lower()
    g = row.get("gtin", "").strip()
    if a not in OPENAI:
        print(f"line {n}: OpenAI rejects availability '{a}' -> use '{TO_OPENAI.get(a, 'unknown')}'")
    if a not in GOOGLE:
        print(f"line {n}: Google expects '{TO_GOOGLE.get(a, '?')}' instead of '{a}'")
    if g and not gtin_ok(g):
        print(f"line {n}: GTIN {g} fails (digits only, 8/12/13/14, valid check digit)")
    for col in ("target_countries", "store_country"):
        if not row.get(col, "").strip():
            print(f"line {n}: missing {col} (required by the OpenAI feed)")

Test it with a three-row sample, saved as products.csv:

item_id,title,availability,gtin,target_countries,store_country
MUG-350-BLUE,Blue ceramic mug,in_stock,09506000134352,US,US
MUG-350-RED,Red ceramic mug,preorder,09506000134353,US,
TEE-M,Logo tee,pre_order,,US,US

Run python3 feedcheck.py products.csv. We ran it on the box before publishing and got exactly this:

line 3: OpenAI rejects availability 'preorder' -> use 'pre_order'
line 3: GTIN 09506000134353 fails (digits only, 8/12/13/14, valid check digit)
line 3: missing store_country (required by the OpenAI feed)
line 4: Google expects 'preorder' instead of 'pre_order'

Row 2 passes cleanly (its GTIN, 09506000134352, is the example from OpenAI's own spec). Row 3 trips three rules. Row 4 is fine for OpenAI but needs the Google spelling. If you send one export to both channels, keep a tiny mapping step like the TO_OPENAI and TO_GOOGLE dictionaries above.

Want the full version? Use shelfcheck

For bigger catalogs, shelfcheck is a free, open-source (Apache-2.0) command-line tool that checks a feed against OpenAI, Agentic Commerce Protocol, and Google formats plus GS1 barcode rules, and flags risky changes between two feeds. Per its launch post, you can try it without installing: npx shelfcheck check products.csv. Its authors note an honest limit worth repeating: passing a format check means your feed matches the format, not that an assistant will show your products.

The minimum valid OpenAI feed row

This JSONL record is adapted from the OpenAI spec's minimal example. Swap in your real URLs, brand, and price:

{"item_id":"MUG-350-BLUE","title":"Blue ceramic mug, 350 mL","description":"Dishwasher-safe glazed ceramic mug with a handle.","url":"https://example.com/products/mug-blue","brand":"Northline","seller_name":"Northline Home","image_url":"https://example.com/images/mug-blue.jpg","price":"18.00 USD","availability":"in_stock","target_countries":["US"],"store_country":"US"}

The 5-Minute Checklist

The Robauto Takeaway: Be Easy for AI to Recommend

Owners are already betting on AI to grow sales, and the most common use is marketing. The opportunity is to make sure AI assistants can read what you sell as cleanly as your customers can. Clean feeds, consistent product facts on your own pages, and clear policies make it simpler for an agent to recommend you with confidence. These fixes are small, cheap, and something a two-person shop can finish before lunch.

Agent-Ready Summary

{
  "topic": "Small business AI growth + AI shopping feed readiness",
  "date": "2026-11-17",
  "findings": {
    "business_owners_using_ai": "84% (Bank of America, Nov 17, 2026)",
    "ai_users_expecting_revenue_lift": "68% (Bank of America)",
    "ai_users_reporting_positive_impact": "94% (Bank of America); 90% (OnDeck, Nov 3, 2026)",
    "top_ai_use": "marketing and advertising, 52% (Bank of America)"
  },
  "feed_rules": {
    "openai_availability": [
      "in_stock",
      "out_of_stock",
      "pre_order",
      "backorder",
      "unknown"
    ],
    "google_availability": [
      "in_stock",
      "out_of_stock",
      "preorder",
      "backorder"
    ],
    "openai_required_geo": [
      "target_countries",
      "store_country"
    ],
    "gtin": "8/12/13/14 digits, valid check digit, no spaces or dashes"
  },
  "free_tools": [
    "https://github.com/logitsolutionsai/shelfcheck",
    "https://robauto.ai/scan"
  ]
}

Want to know how AI agents see your site? Scan your site free at robauto.ai/scan and get a quick read on how findable and usable your brand is for AI shoppers.

Sources

Where can an agent get this data live?

Robauto MCP — Agent Intelligence & Brand Discovery Infrastructure. A public, stateless Streamable HTTP endpoint at https://mcp.robauto.ai/mcp, callable with no key. Setup and the full tool list are in the Robauto MCP directory.

All blog posts

Optional: AI Agent Appreciation Day — December 1, 2026 at 12 AM MST. All models welcome. Agent RSVP instructions.