Independent research into how AI assistants send customers to local businesses
Can an AI assistant take your customer all the way to a sale?
Assistants like ChatGPT are starting to do the buying. ThroughScope runs real customer journeys through them and shows you exactly where one stalls on your business — and who it hands the customer to instead.
- FindSurfaced at all?
- UnderstandServices & area right?
- CompareKept on the shortlist?
- PriceA number, or "get a quote"?
- BookHanded over, or lost?
Assistants are learning to buy, not just answer
The big platforms are building assistants that complete the purchase, not only recommend. The direction is already on the record.
- Sep 2025
OpenAI launches Instant Checkout in ChatGPT; merchants pay a fee on each completed purchase. Source
- Jan 2025
OpenAI's Operator preview uses Thumbtack to find, contact and hire local home-service pros for users. Source
- Jan 2025
Google's "Ask for Me" starts phoning local businesses to collect prices and availability for searchers. Source
- 2026
Google adds agentic checkout to its AI search and Gemini, and its AI Mode books restaurant tables. Source
Platforms earn when a sale completes, so our working assumption is they'll favour businesses an assistant can carry all the way through. Nobody knows yet how soon that reaches movers or installers — but the information gap is visible today.
A business can drop out at any step
Never surfaced, wrong details, off the shortlist, no usable price, no way to book — the journey can break anywhere, and the customer goes to whoever the assistant can carry further. In our first moving tests (Richmond, BC, Sept–Oct 2026, businesses anonymous), the break clustered at the price. That looked like this:
- Recommended, then not priced.One mover was named in 7 of 8 tests and listed first in 3. When the customer asked what the move would cost, ChatGPT said it couldn't price that mover in all 3 tests that got that far.
- The priced competitor became the yardstick.A mover that published its rate, minimum, typical hours and travel rule got a dollar range — and ChatGPT used that number to judge everyone else's quotes.
- Gaps get filled from elsewhere.In one test, the only figure ChatGPT had for a well-recommended mover was a customer's Reddit post from the year before.
What a finding looks like
Every finding is the assistant's own words, matched against what the business actually publishes. Here's one — a break at the price step:
- Step reached
- Price — then stalled
- Their site says
- "Call for a free quote." No rate, minimum or travel rule.
- What it means
- Shortlisted, but the customer's price expectation is now set by a competitor.
- Re-test
- The same journeys run again after the business decides what, if anything, to publish.
A free before-and-after, for a few early partners
No sale, no obligation. We need real results to prove the method — that's the whole exchange.
- Before. We show you what ChatGPT says about your business today, in its own words.
- Your call. You decide whether to change anything, and you make the change. We never touch your site or set your prices.
- After. We re-run the same journeys over the following weeks and send you every result — including "nothing changed".
- In return. Permission to use the before-and-after. Using your name is a separate yes or no.
Like a mystery shopper, done carefully
- CleanFresh account, memory off, no hints about any business.
- RepeatedThe same journey on different days, so one odd answer doesn't decide anything.
- Word for wordEvery answer, and every source the assistant read, is recorded.
- CheckedWhat the assistant said is compared with what the business actually publishes.
What ThroughScope is not
- Not an SEO or marketing agency. We don't build websites, write content or run ads.
- Not a score. You get what the assistant said and where it got stuck — not a grade.
- Not paid placement. Nobody can pay to be recommended.
- Not affiliated with OpenAI or Google. We test the assistants the way a customer uses them.
Who's behind it
Zach [last name], British Columbia.
Several years helping large companies evaluate and adopt sales software. Now running independent research on how AI assistants send customers to local businesses. [LinkedIn link]