Luna Studios Research · Edition 05
The Talking
Web 2026.
How much of American business can answer today — and what that says about the next ten years.
The Speaking Index 2026 · United States
0.05
out of 100
The share of measured business websites that can answer a visitor on their own. One in 2,186 American business websites is able to do it.
01 · The behavioral shift
It took three years to retrain
a billion people.
People who want to know something no longer type keywords into a box and work through a list of results. They ask a question — in full sentences, with context, with follow-ups — and expect an answer. Not ten blue links. An answer.
That habit is not confined to one service. It transfers. Anyone who has learned to ask instead of search carries that expectation everywhere — including onto the website of a contractor, a dental practice, an accounting firm.
The expectation changed. The websites did not.
This report measures how wide the gap has become. Not as an opinion, but as a number — updated annually, using the same method as every other edition in this series.
02 · What people expect
Patience is no longer a currency.
The expectation side is well researched — it is just rarely gathered in one place. The most solid findings from independent studies:
And the expectation keeps tightening: for AI-driven applications, 54 % already expect an answer within seconds. What counted as fast service yesterday is the floor today.
The cost of waiting has been quantified precisely. A study by MIT and InsideSales analyzed more than 15,000 enquiries: responding within five minutes instead of thirty makes you 100 times more likely to reach the prospect and 21 times more likely to qualify them. An audit of 2,241 companies published in Harvard Business Review found the same pattern: firms replying within an hour were seven times more likely to qualify a lead than those replying an hour later — and sixty times more likely than those who waited a day.
That same audit surfaced a second finding that is still underrated: 23 % of the companies studied never responded at all. Not late. Never.
03 · What websites deliver
We asked 275,439 American
business websites a question.
126 could answer.
The basis is 377,198 US businesses with their own website across all 50 states — restaurants, medical practices, contractors, hotels, salons, law firms, schools, studios. Each website was requested once in August 2026 and the delivered source code checked for 30 common chat, messaging and assistant systems. 275,439 websites were reachable; all figures below refer to those.
Of the 275,439 reachable websites, 2,562 had any chat or assistant system at all — 0.93 %. The vast majority of those are conventional chat windows that only answer when a human is sitting behind them. Systems capable of answering a question on their own were found on 126 websites. That is 0.05 %.
One in 2,186 American business websites can answer a visitor on its own.
Because the boundary can be argued, we report both readings: counting only systems whose integration is unambiguously present in the source code gives 0.93 %. Counting every trace of such a system — including analytics suites that merely could offer chat — gives 2.10 %. Either way, it stays far below the threshold of visibility.
04 · By state
A country of many languages.
Websites in one.
On answering behavior the states are almost indistinguishable — the spread is a fraction of a percent at a level below one percent. On multilingualism something more striking appears: not a single state reaches ten percent. California, Florida, Arizona and Texas — states where Spanish is spoken in millions of households — sit barely above the national average of 5.6 %.
| State | Websites | Reachable | Contact form | Multilingual |
|---|---|---|---|---|
| District of Columbia | 1,925 | 71.8 % | 45.4 % | 9.3 % |
| North Dakota | 508 | 76.7 % | 53.5 % | 8.1 % |
| Florida | 10,969 | 69.9 % | 43.4 % | 7.6 % |
| California | 29,863 | 71.5 % | 42.0 % | 7.2 % |
| Arizona | 4,657 | 71.7 % | 42.1 % | 7.0 % |
| Alaska | 1,169 | 77.5 % | 43.2 % | 6.8 % |
| Mississippi | 903 | 69.5 % | 45.1 % | 6.8 % |
| New York | 20,351 | 72.2 % | 39.5 % | 6.8 % |
| Hawaii | 1,293 | 71.4 % | 37.5 % | 6.7 % |
| Alabama | 2,849 | 70.2 % | 44.7 % | 6.6 % |
| Rhode Island | 994 | 74.6 % | 44.5 % | 6.3 % |
| Nebraska | 1,915 | 72.0 % | 44.1 % | 6.3 % |
| Texas | 15,475 | 72.7 % | 41.1 % | 6.2 % |
| Illinois | 12,185 | 75.2 % | 42.9 % | 6.2 % |
| Idaho | 1,704 | 74.2 % | 36.8 % | 6.2 % |
| Connecticut | 3,786 | 73.4 % | 45.5 % | 6.2 % |
| Nevada | 1,873 | 71.1 % | 37.6 % | 6.2 % |
| Delaware | 885 | 77.8 % | 45.8 % | 5.9 % |
For comparison: in Switzerland 22.8 % of business websites offer a second language, in Austria 16.7 %, in Germany 8.2 %. The United States, by far the most linguistically diverse of the four markets measured, has the lowest share at 5.6 %. This is not a question of capability — translation has never been cheaper. It is a question of who a website assumes it is talking to.
05 · By industry
No industry is ahead.
Industry categories come from the underlying data source, so they are recorded rather than estimated. The picture is remarkably uniform across every group.
| Industry | Websites | Contact form | Tappable phone | Multilingual | Instant channel |
|---|---|---|---|---|---|
| Restaurants | 42,795 | 28.7 % | 42.7 % | 4.0 % | 0.30 % |
| Places of worship | 19,215 | 38.6 % | 23.7 % | 5.0 % | 0.18 % |
| Schools | 14,930 | 48.5 % | 43.3 % | 6.4 % | 0.47 % |
| Cafés | 7,362 | 33.1 % | 23.0 % | 3.5 % | 0.42 % |
| Dental practices | 4,875 | 37.5 % | 77.5 % | 4.3 % | 1.91 % |
| Museums | 4,809 | 51.5 % | 27.9 % | 5.8 % | 0.25 % |
| Bars | 4,471 | 31.6 % | 30.7 % | 3.7 % | 0.34 % |
| Fast food | 4,325 | 26.8 % | 32.3 % | 4.0 % | 0.46 % |
| Corporate offices | 4,185 | 45.5 % | 41.5 % | 12.5 % | 2.56 % |
| Hair salons | 4,122 | 24.6 % | 31.8 % | 3.7 % | 0.46 % |
| Golf Course | 3,941 | 47.9 % | 53.9 % | 7.6 % | 0.23 % |
| Auto repair shops | 3,655 | 30.9 % | 59.1 % | 4.6 % | 0.82 % |
One number deserves attention. Among 4,875 dental practices, 1.91 % offer an instant channel — the highest value of any large category in this study, and more than five times the national average. The same measurement in Germany found zero among 5,297 dental practices. Two comparable professions, two continents, opposite ends of the scale. Whatever is holding this back, it is not the nature of the work.
06 · What the pages say
The website forces the question —
and offers no way to ask it.
The broad measurement tells us whether a website can answer. It says nothing about whether an answer would be needed in the first place. So we additionally analyzed a deep sample of 400 randomly drawn businesses — for these, not just the homepage but up to six subpages were read in full.
Only together do these produce the full picture. The website does not answer the most obvious questions — what it costs, how it works, when someone is there. It therefore forces the visitor to ask. And then it offers no way to do so except the telephone.
07 · The opening hours of the internet
The website is awake
168 hours a week.
The business is awake 45.
Among the deep-sample websites that state specific hours, the picture is consistent: availability begins at a median of 9:00 and ends at 18:00.
That leaves roughly 73 % of the week in which the website is alone. And those are not the fringes of business: evenings at the kitchen table, weekends, lunch breaks — precisely when people compare providers, read offers and make decisions. During those hours the website is the only contact a company has with the world. And it says nothing.
08 · The gap
Two sentences that do not fit together.
What people expect
- 83 % want to interact with someone immediately
- More than half expect an answer within an hour
- 54 % expect seconds from AI-driven services
- Questions arise at night and on weekends, not during office hours
- Whoever gets no answer asks the next company
What websites offer
- 99.1 % have no channel for an immediate answer
- 0.5 % commit to when an answer will come
- 0.05 % can answer on their own
- 45 of 168 weekly hours someone is there
- 65.8 % do not even state a price
The gap is not incremental. It is structural: one side learned to ask. The other never learned to answer.
09 · What this means for the United States
This is not the exception.
This is the norm.
The United States is home to over 33 million businesses, 99.9 % of them small and mid-sized. The 377,198 businesses measured here are typical of that group: owner-operated, with a maintained website, without an in-house IT department.
Applied to that population, the finding describes the default state: reachable during office hours, silent for the remaining 73 % of the week, with no way for a visitor to get a simple question answered.
Every unanswered question is a buying signal that evaporates. Not rejected — merely unanswered.
What that costs cannot be stated responsibly as a single dollar figure; industries and order sizes differ too much. What can be said: the research on response time is unambiguous, it is decades old, and it has been confirmed repeatedly. The difference between an answer in five minutes and an answer the next day is not a comfort feature. It is the difference between a customer and a statistic.
10 · Outlook
Why this number
only moves one way.
The internet has changed fundamentally twice, and both times it went the same way. In 1993 the browser arrived, and companies asked why they would need a website. In 2007 the smartphone arrived, and the same question was asked about mobile layouts. Both times the share of early adopters started in the low single digits. Both times, within a few years, the question was no longer whether — but why not yet.
The Speaking Index stands at 0.05 in 2026. It is a number that did not exist before, because there was nothing to measure. And it is a number that becomes interesting in hindsight: anyone counted in it today belongs to 126 businesses in the entire United States.
11 · Test your own
Where does your website stand?
The most interesting number in this report is the one that is not in it: yours. Enter your domain — Luna, the assistant on this page, will look at the website in real time and tell you which of the questions studied here it can answer and which it cannot.
The check runs as a conversation with Luna. Nothing is stored and nothing is sent.
Scan the code or open the address, enter a domain — Luna inspects the website in real time and reports which of the questions studied here it can answer.
Method
How this was measured.
- Population. 377,198 US businesses with their own website, drawn from openly licensed geodata (OpenStreetMap, ODbL) across all 50 states. Included are businesses recorded there with their own web address; links to platform profiles such as social networks or directories were excluded. This is not a random sample of the entire US business landscape — businesses without a website and businesses absent from that data are missing. The real state of affairs is therefore likely worse rather than better.
- Live measurement. Each website was requested exactly once in August 2026 and the delivered source code checked for 30 common chat, messaging and assistant systems. 275,439 of 377,198 websites were reachable; 46,079 did not respond, 55,680 refused automated access.
- Verification and correction. Every hit was reloaded individually and checked against strict detection patterns. Loose patterns produce substantial false positives — through word fragments in foreign-language text, or through analytics tools that share an address with a related chat function. Only confirmed hits enter the figures. In this edition, 3,217.0 of the initial hits were discarded as false positives.
- Two readings. Strictly counted, a system registers only when its integration is unambiguous in the source code (0.93 %). Loosely counted, every trace registers (2.10 %). This report uses the strict reading.
- Limits. Only the delivered source code is examined. Systems injected later by JavaScript are invisible to this method — the figures are therefore lower bounds. Websites behind bot protection were not counted as lacking an answer channel but reported separately.
- Speaking Index. The share of websites able to give a visitor a substantive answer without human involvement. 2026 United States: 0.05 (126 of 275,439).
- Deep sample. Figures on prices, frequently asked questions, hours and response commitments come from 400 randomly drawn websites read in full, including up to six subpages (median 4 pages). Hours could be parsed unambiguously on 138 websites; medians are reported.
- Vendors. Individual chat and assistant products are deliberately not named. The subject of this study is whether a website can answer — not which product it uses.
- Comparability. Method, detection patterns and analysis are identical across every edition in this series, including the German-language editions for Switzerland, Germany and Austria. Country figures are therefore directly comparable.
- Third-party data. All figures on the expectation side come from independent studies by others. Luna Studios conducted no consumer survey for this report.
Sources
Oldroyd, McElheran, Elkington: "The Short Life of Online Sales Leads"Harvard Business Review 89 (3), March 2011. Audit of 2,241 companies. hbr.org
Oldroyd et al.: Lead Response Management StudyMIT / InsideSales, 2007. Analysis of more than 15,000 enquiries. The 100× and 21× multipliers frequently attributed to Harvard Business Review originate from this study.
Expectations of response times in customer contactCompilation of current studies on service speed and digital patience, including the Emplifi Consumer Survey. Overview
U.S. Small Business Administration: Small Business ProfilesThe United States is home to over 33 million businesses, 99.9 % of them small and mid-sized. Source
Business data© OpenStreetMap contributors, licensed under the Open Database License (ODbL). openstreetmap.org/copyright
Own measurementLuna Studios, August 2026. 377,198 websites, 275,439 measured live, every hit verified individually.
Use
Citation explicitly welcome.
Every figure and chart in this report may be used freely — in articles, talks, studies and internal documents. We ask for attribution:
Luna Studios: "The Talking Web 2026 — United States", lunastudios.ai/research
Who publishes this report
Luna Studios builds Luna — an AI assistant with a real voice that lives on company websites and answers visitors around the clock. We have an obvious interest in the subject this report measures. That is precisely why every figure on the expectation side comes from third-party sources, why the measurement is disclosed in full — including the false positives we corrected along the way — and why this report contains no product claim of any kind.