When a small-business owner in the northwest suburbs asks AI who can help them with AI, Syndesi appears in 0 of 77 measured answers.

2. What we did, and what the numbers are out of

We built 33 real buyer questions across five types: what-is and how-much questions, Chicago-area questions, town-by-town questions for eleven northwest suburbs, industry questions, and how-does-it-work questions. We asked all 33 to ChatGPT, Claude, Perplexity, and Gemini, three times each, with live web search on for every counted answer. Every answer was recorded in full, and every number on this page can be checked against a stored answer.

Two honest notes about the counting, because we would rather show a smaller true number than a bigger padded one.

First, the headline number covers the 21 questions where a local business could be the answer. Those 21 questions across four systems make 84 possible answers, and 77 could be measured to our standard, so the headline is 0 of 77 and every competitor count on this page is out of the same 77.

Second, not every question could be measured on every system. Some questions are general enough that the AI answers from what it already remembers instead of searching the live web, and an answer like that is not what your customer sees, so we set it aside and say so instead of guessing. Across all 33 questions and four systems, that came to 107 measured answers, 22 set aside for answering without searching, and 3 set aside because the question needs to know where you are and that system cannot be told a location. The near-me question can only be measured properly on ChatGPT; we state that rather than working around it.

We also checked the instrument against the real apps. We drove the actual consumer apps, in fresh sessions with memory off, asked a sample of the questions exactly as written, and compared. Our results matched the real app in 16 of 21 comparable answers. One of the five differences came from asking the near-me question from a neighboring town to the one the instrument used, and we say that rather than hide it. The check also caught our own instrument under-counting one competitor; we fixed the instrument and re-ran every number on this page before publishing. We publish the match rate because an owner who asks "does this reflect what my customer actually sees" deserves a number, not a reassurance.

3. The scoreboard

Every business we follow month to month, across all four systems, out of the same 77 measured answers.

"Named sometimes" means the system named them in one answer out of three for the same question. Answers move between askings, so a one-of-three name is noted but never counted. Syndesi has zeros in both columns: not named once, in any answer, on any system.

One note on source links: when an answer links to a business's own website, we track that separately as the stronger signal. Google's system reports its sources through its own redirect links, so we can rarely confirm which business a Gemini link belongs to, and where we could not confirm, we did not count it.

4. Where we show up, and where we are invisible

By question type. "Measured" counts answers that met our standard; the rest are stated, not guessed.

Syndesi appeared in none of the categories. The whole board, not one square, and every square is empty.

Two things worth seeing in this table. The town-by-town rows, the heart of local visibility, produced 44 clean measured answers and named our competitors freely. And the what-is and how-much rows, the questions a curious owner asks first, mostly got answered from the AI's memory rather than a live search, which means nobody local is winning those answers yet. They are sitting there unclaimed.

5. Who gets named instead of us, and why

First, the finding that reframes everything: the most-named business on any local question in this run is not local. Zfort Group, a software firm with no local presence, was named in 5 of 77 measured local answers and linked as a source in all 5. It ranks because it publishes a town-name page for city after city, and on some questions it is the only business the AI can find at all. Five more businesses cleared our counting bar the same way and were confirmed non-local on review: national and out-of-metro firms surfacing through town-name landing pages. When the front-runner for "who is local" is a company that is not, the lane is open.

GO Technology Group (named in 16 of 77). An IT services firm with offices in Oak Brook and Chicago. The stored answers show exactly why they win: their pages list town after town by name, Arlington Heights and Des Plaines among them, so when a question names a suburb, there is something specific for the AI to find. Named on Claude, Gemini, and Perplexity; notably, the answers named them 16 times without once linking to their site. Being findable in third-party listings is carrying them.

Expert Management Group (named in 6 of 77, linked as a source in 4). A Schaumburg firm we had never tracked, surfaced by reading every answer rather than checking a list. The answers describe them targeting small-business AI needs directly: automation, chatbots, practical integrations.

ProdigyTeks (named in 6 of 77, linked in 3). A Chicago IT support company whose content speaks directly to small-business buyers, and it shows: one answer named them the safest starting point for a small business in Illinois.

Xcelacore (named in 5 of 77, linked in 3) and CelereTech (named in 4 of 77, linked in 4). Oak Brook and Chicagoland firms whose pages state their service areas town by town. CelereTech gets named for Elk Grove Village because its site says, in plain text, that it serves Elk Grove Village. That is the whole trick.

The longer list. In total, 25 confirmed local competitors were named across the 77 answers, and 22 of them were businesses nobody thought to track before this run found them. Beyond those, more than 100 further businesses were named a single time and never again; answers churn heavily between askings, so a one-time name is noise, not standing. Consistency is what separates the businesses above from the crowd.

And our tracked list from July. Ademero, the Florida document company with template town pages, has faded to 3 of 77 this run but still gets named for Chicago-area questions it has no presence in. nvisia was named twice, both on town questions. Quantum Rise once, on an industry question. The credentialed downtown firms are barely in this conversation; the winners are the ones with something town-specific to find.

6. What AI gets wrong about us

Nothing, for the least comforting reason possible. Across every measured answer in this run, 319 of them, Syndesi was named zero times. That is a real zero, not a technicality: every one of those answers came from a live search that could have named us and did not. AI says nothing false about Syndesi because it says nothing about Syndesi at all.

7. What the real apps show

The screenshots below come from the actual consumer apps, captured August 23, 2026, in fresh sessions with memory off, asking the questions exactly as our instrument does.

8. The fix list

The same list we would hand a client. We are simply our own first client, and we grade ourselves in public.

1. 1.Publish this report. Done. You are reading it. A full, honest answer to what an AI visibility assessment shows, with real data on our own domain, zero included.
2. 2.Build the third-party base. Chamber of commerce listings in Arlington Heights, Mount Prospect, and Schaumburg, plus Clutch, Better Business Bureau, Apple Business Connect, Alignable, and Nextdoor. Section 5 is the receipts for why: the winners are the ones findable in places other than their own site. In progress this week.
3. 3.Say the towns, and answer the questions. One real page per buyer question, starting with the what-is and how-much questions nobody local is winning, with our service area stated in plain text the way CelereTech states Elk Grove Village. Rolling out, starting with this page.
4. 4.Machine-readable business info on syndesi.io. Our name, services, prices, and service area structured so AI systems can read them without guessing. Done, August 10, 2026.
5. 5.Re-measure monthly and publish the chart. The same questions, every month, same method, chart at the top of this page. The climb from zero is the ongoing proof the product works.

9. What happens next

For us: the list above, measured monthly, published openly, starting from zero.

For you: this is the report. Yours costs $149 and comes with a page-by-page walkthrough where you can ask anything you want. If you want us to fix what it finds, we quote that separately once you have seen what is needed. The $149 buys the report, not a sales pitch.