2026 Q4 · Version 1 · Published October 12, 2026 · Research by Modern Mustard Seed
The Small Business AI Readiness Index
When a customer asks ChatGPT, Claude or Google for a plumber, a dentist or a bakery, can the assistant find the business, read it and trust it? I wanted a number instead of a hunch, so I measured 2,000 small business websites drawn at random from OpenStreetMap across all 50 states and DC, with the same checks we publish as free connectors for Claude.

Headline findings
What 2,000 sites told us
- 58.5
The average small business website scores 58.5 out of 100.
Each site is scored on eight pass or fail checks that decide whether a customer asking an AI assistant can find it, read it and trust it. Most sites pass the basics: HTTPS, open to AI agents, a phone and a way to get in touch. The same few checks pull the average down.
95% CI 57.4 to 59.7 · base: 1,830 sites with at least 5 of the 8 checks measurable
- 11.3%
Only 11.3% enforce DMARC on their email domain.
DMARC at quarantine or reject is the setting that stops a stranger sending an invoice in the business name. It is the weakest of the eight checks. Another 21.5% publish DMARC but only monitor.
95% CI 9.9 to 12.8 · base: 1,969 sites on their own domain
- 24.7%
Only 24.7% carry valid LocalBusiness markup.
LocalBusiness structured data tells a machine the name, address, hours and phone without guessing. 32% have some; the difference is markup that fails validation.
95% CI 22.5 to 27.1 · base: 1,512 sites whose homepage loaded
- 47%
47% of websites sit on a platform that is hard or locked to move.
Where we could rate the website platform, nearly half would be hard to leave with the site intact. Domains look well kept, with 92.7% carrying a transfer lock. The lock-in sits in the website layer.
95% CI 43.9 to 50.2 · base: 1,065 sites whose homepage loaded and whose platform was rated
- 1.3%
Access is not the gap: 1.3% block an AI search agent.
Almost every small business lets ChatGPT, Claude and Perplexity read the homepage. What those agents find when they arrive is the problem.
95% CI 0.8 to 1.9 · base: 1,740 sites with a readable robots.txt answer
- 17%
17% of listed websites did not load at all.
DNS failures, refused connections, timeouts, TLS failures and error pages. A customer, or an AI assistant sent on their behalf, meets the same dead end.
95% CI 15.2 to 18.8 · base: 1,830 sites whose robots.txt let us read the homepage and that did not refuse automated readers
The Index measures what a business controls on its own site, its email domain and its registration. It does not measure how often any AI product names a business, and it makes no claim about placement.
The Index in one view
Eight checks, one number
Each site earns a point for each of eight checks that apply to it, and its score is the share passed, out of 100. The Index is the average of those scores with national weights. The weakest link nationally is DMARC enforced, passed by 11.3% of the sites it applies to.
Who may read the site
AI agents at the door
1.3% of sites with a readable robots.txt block at least one AI search or AI assistant agent, and 0.9% block every one we checked. 1.9% opt out of at least one training crawler, which OpenAI, Google and Apple document as separate from their search products.
Can a machine and a customer tell who you are
Structured data and contact paths
Trust: inboxes, browsers and registries
The quiet gaps
By trade
Readiness by trade
| Trade | Sites | Index score | Blocks an AI search agent | Valid LocalBusiness | Booking path | DMARC enforced |
|---|---|---|---|---|---|---|
| Health, dental and veterinary | 176 | 6662 to 69 | 1%0 to 4 | 34%26 to 42 | 50%42 to 58 | 13%9 to 19 |
| Professional services | 216 | 6360 to 67 | 2%1 to 5 | 14%9 to 20 | 21%16 to 28 | 24%19 to 30 |
| Trades and home services | 138 | 6157 to 66 | 0%0 to 3 | 28%20 to 37 | 31%23 to 40 | 17%11 to 24 |
| Lodging, fitness and other services | 154 | 5853 to 62 | 4%2 to 9 | 23%16 to 31 | 38%30 to 48 | 12%8 to 19 |
| Retail shops | 266 | 5754 to 60 | 2%1 to 5 | 20%15 to 26 | 13%9 to 18 | 9%6 to 13 |
| Food and drink | 692 | 5755 to 59 | 1%0 to 2 | 27%24 to 31 | 50%46 to 54 | 7%6 to 9 |
| Beauty and personal care | 98 | 5549 to 60 | 1%0 to 6 | 24%16 to 35 | 51%40 to 62 | 5%2 to 12 |
| Auto and vehicle services | 90 | 5347 to 59 | 0%0 to 5 | 30%20 to 42 | 14%7 to 25 | 11%6 to 20 |
By state
Readiness by state
State samples are small, so states are listed alphabetically, not ranked. Read the interval under each figure.
| State | Sites | Index score | Blocks an AI search agent | Valid LocalBusiness | DMARC enforced |
|---|---|---|---|---|---|
| Alabama | 20 | 6250 to 73 | 0%0 to 16 | 35%17 to 59 | 9%2 to 27 |
| Alaska | 14 | 5643 to 70 | 0%0 to 24 | n under 10 | 13%4 to 36 |
| Arizona | 32 | 6661 to 71 | 0%0 to 10 | 13%5 to 29 | 15%6 to 30 |
| Arkansas | 19 | 5742 to 71 | 0%0 to 18 | 27%11 to 52 | 11%3 to 31 |
| California | 156 | 5955 to 62 | 0%0 to 3 | 28%21 to 36 | 13%9 to 20 |
| Colorado | 53 | 6458 to 69 | 2%0 to 10 | 26%16 to 40 | 20%12 to 32 |
| Connecticut | 23 | 5543 to 68 | 0%0 to 16 | 24%10 to 47 | 15%6 to 34 |
| Delaware | 10 | 4932 to 67 | 0%0 to 23 | n under 10 | 0%0 to 19 |
| District of Columbia | 19 | 5442 to 67 | 0%0 to 18 | 27%11 to 52 | 14%5 to 33 |
| Florida | 65 | 5953 to 65 | 3%1 to 12 | 30%19 to 44 | 18%11 to 29 |
| Georgia | 40 | 6254 to 71 | 0%0 to 9 | 29%17 to 46 | 16%8 to 29 |
| Hawaii | 19 | 5544 to 65 | 0%0 to 18 | 31%14 to 56 | 5%1 to 24 |
| Idaho | 18 | 6352 to 74 | 0%0 to 18 | 27%11 to 52 | 5%1 to 24 |
| Illinois | 65 | 6155 to 66 | 0%0 to 6 | 20%11 to 32 | 15%8 to 26 |
| Indiana | 27 | 5745 to 69 | 0%0 to 14 | 38%21 to 59 | 10%4 to 26 |
| Iowa | 24 | 5845 to 71 | 0%0 to 14 | 33%16 to 56 | 12%4 to 29 |
| Kansas | 24 | 5139 to 63 | 0%0 to 18 | 24%10 to 47 | 12%4 to 30 |
| Kentucky | 22 | 3421 to 47 | 0%0 to 23 | 9%2 to 38 | 5%1 to 23 |
| Louisiana | 22 | 6048 to 73 | 5%1 to 22 | 33%16 to 56 | 13%4 to 31 |
| Maine | 18 | 6049 to 71 | 0%0 to 16 | 20%7 to 45 | 18%7 to 39 |
| Maryland | 45 | 6053 to 67 | 0%0 to 8 | 18%9 to 33 | 26%16 to 40 |
| Massachusetts | 41 | 5446 to 63 | 2%0 to 13 | 26%14 to 43 | 4%1 to 14 |
| Michigan | 51 | 5851 to 65 | 0%0 to 7 | 36%23 to 51 | 9%4 to 20 |
| Minnesota | 40 | 6459 to 69 | 5%1 to 17 | 14%6 to 28 | 5%1 to 16 |
| Mississippi | 14 | 7056 to 84 | 0%0 to 20 | 39%18 to 65 | 19%7 to 43 |
| Missouri | 31 | 6356 to 71 | 0%0 to 11 | 29%15 to 47 | 12%5 to 27 |
| Montana | 17 | 6356 to 71 | 0%0 to 18 | 13%4 to 36 | 24%10 to 47 |
| Nebraska | 19 | 6152 to 70 | 0%0 to 18 | 29%13 to 53 | 5%1 to 24 |
| Nevada | 21 | 5546 to 65 | 0%0 to 18 | 20%7 to 45 | 14%5 to 33 |
| New Hampshire | 25 | 4936 to 62 | 0%0 to 16 | 24%10 to 47 | 7%2 to 23 |
| New Jersey | 39 | 5647 to 65 | 3%1 to 16 | 40%25 to 58 | 5%1 to 17 |
| New Mexico | 20 | 5440 to 68 | 6%1 to 28 | 25%10 to 50 | 20%8 to 42 |
| New York | 106 | 6157 to 65 | 3%1 to 8 | 23%15 to 32 | 9%5 to 16 |
| North Carolina | 50 | 5650 to 62 | 0%0 to 8 | 22%12 to 37 | 8%3 to 19 |
| North Dakota | 10 | 5031 to 69 | 0%0 to 26 | n under 10 | 29%12 to 55 |
| Ohio | 48 | 5447 to 62 | 0%0 to 9 | 20%10 to 36 | 14%7 to 26 |
| Oklahoma | 20 | 6554 to 77 | 0%0 to 17 | 22%9 to 45 | 14%5 to 35 |
| Oregon | 63 | 5750 to 63 | 4%1 to 12 | 25%15 to 38 | 10%5 to 20 |
| Pennsylvania | 67 | 6055 to 66 | 2%0 to 8 | 27%17 to 39 | 4%1 to 11 |
| Rhode Island | 15 | 5239 to 65 | 0%0 to 23 | 17%5 to 45 | 6%1 to 28 |
| South Carolina | 34 | 5748 to 66 | 0%0 to 12 | 18%8 to 36 | 14%6 to 29 |
| South Dakota | 16 | 6755 to 79 | 0%0 to 20 | 33%15 to 58 | 25%10 to 50 |
| Tennessee | 25 | 5748 to 66 | 4%1 to 19 | 15%5 to 36 | 7%2 to 22 |
| Texas | 82 | 5751 to 62 | 0%0 to 5 | 23%14 to 34 | 5%2 to 11 |
| Utah | 29 | 6255 to 70 | 0%0 to 12 | 19%8 to 37 | 13%5 to 29 |
| Vermont | 18 | 5644 to 68 | 0%0 to 18 | 40%20 to 64 | 5%1 to 25 |
| Virginia | 60 | 6053 to 66 | 4%1 to 13 | 22%13 to 35 | 8%4 to 18 |
| Washington | 60 | 5549 to 61 | 2%0 to 9 | 23%13 to 37 | 9%4 to 18 |
| West Virginia | 15 | 6251 to 73 | 0%0 to 22 | 21%8 to 48 | 7%1 to 32 |
| Wisconsin | 43 | 6255 to 68 | 2%0 to 12 | 26%15 to 42 | 18%10 to 32 |
| Wyoming | 16 | 5236 to 68 | 0%0 to 24 | 27%10 to 57 | 13%4 to 36 |
Methodology
How I measured it
Population and frame
Independent US small businesses with a website. I read every OpenStreetMap shop, craft and office feature, plus selected amenity, healthcare, leisure and tourism features, in the 50 states and DC that list a website (OpenStreetMap snapshot 2026-10-12): 642,707 features. Out came institutions and government offices, anything tagged with a brand (chains), listings whose website is a social or directory page, private addresses, and any domain listed on three or more features (multi-location). The frame holds 251,025 small business websites. OpenStreetMap was chosen because its license allows this use; commercial directories forbid it.
Sample
2,000 sites, stratified by state with a floor of 12 per state and the rest in proportion to each state's frame, spread across eight trades by a seeded systematic draw (seed 4060752512). Every site carries the design weight of its state, and national and trade estimates are weighted back to the frame.
Measurement
The engines of four of our connectors for Claude (Site Check, AI Search Kit, Domain Health and Own It, engine commit 4c3515ef3e99d83571d2cdc7ed535f93694281c4) read each site's robots.txt, homepage and /llms.txt, its public DNS, one TLS handshake and its registry record. robots.txt is read first and obeyed for our crawler, MMSReadinessIndex. One fetch per resource, no forms submitted, no logins, no inner pages.
Outcomes
All 2,000 sampled sites were measured. 1,512 homepages loaded, 318 did not, 134 refused automated readers and 36 excluded our crawler in robots.txt. Every metric names its own base in the CSV, and a check that could not be measured is never counted as a failure.
Estimates
Shares are weighted proportions with 95% Wilson intervals on the effective sample size. The Index score is a weighted mean with a linearised interval. No estimate is published for a group under 10 sites, and groups under 30 are flagged in the CSV.
Limits
Homepage only, so phone, booking and schema on inner pages are not seen and those shares are floors. DKIM is a floor. OpenStreetMap covers cities better than contractors who work from a truck, and version 1 is not weighted to Census counts. One read per site: a site down that minute counts as not loading.
The full method, the eight-check definition and every metric's denominator are in the PDF report and the CSV.
Privacy
Aggregates only. No business is named.
No business name, domain, URL or per-site score appears on this page, in the CSV or in the report, and the publishing job fails if any sampled domain shows up in public output. Nobody was contacted because of this study.
Downloads and license
Take the data
- The report, 2026 Q4 (PDF, 5.4 MB)Findings, charts, trade and state tables, and the method.
- The aggregate data, 2026 Q4 (CSV, 248 KB)Every published estimate with its base, n, effective n and 95% interval, for the nation, 8 trades and 51 states.
Business listings © OpenStreetMap contributors. The aggregates are a derived database of OpenStreetMap and are released under the Open Database License 1.0: use them, share them and build on them with attribution, and keep derived databases open.
Cite this
Quote it, link it
Please cite the Index and link this page, so readers can check every number against its interval and base.
Modern Mustard Seed. (2026). Small Business AI Readiness Index, 2026 Q4 (Version 1) [Data set]. Published October 12, 2026. https://modernmustardseed.com/readiness-index
@misc{mms_readiness_index_2026q4,
author = {{Modern Mustard Seed}},
title = {Small Business AI Readiness Index, 2026 Q4},
version = {1},
year = {2026},
month = oct,
note = {Aggregate data under the Open Database License 1.0. Business listings (c) OpenStreetMap contributors},
url = {https://modernmustardseed.com/readiness-index}
}Press
Questions about the numbers?
I answer press questions myself. Email sarah@modernmustardseed.com for a cut of the data by trade or state, the charts at full size, or the method behind any figure.
Sarah Scarano, founder, Modern Mustard Seed
Your own business
Run the same checks
Every check in the Index comes from a free connector anyone can add to Claude. Ask it whether your email can be faked, whether AI search can read your site, or who controls your domain, and get the fix with the answer.
Claude is a product of Anthropic. Modern Mustard Seed is an independent studio and is not affiliated with or endorsed by Anthropic.