Home»Client results»WaitWell
AI search booking demos WITHIN 2 months
How WaitWell turned AI search into booked demos and revenue.
Buyers were shortlisting inside AI assistants, then booking. Four months in, demos and new customers from search and AI are running at 2.9× and 3.3× the forecast we published before kickoff.
The clientQueue management, for places people wait.
WaitWell builds queue management and appointment scheduling software for healthcare, government, education and service counters, replacing physical lines with virtual queues, kiosks and check-in flows.
It is a category where buyers shortlist by comparison: best-of lists, alternatives pages and head-to-head matchups, increasingly read inside an AI assistant rather than a search results page.
The problemNearly half the pipeline had no source.
WaitWell was booking demos. What nobody could say was where they came from.
The CRM classifier filed 42% of demo bookers as Direct Traffic, the bucket that means the system does not know, and only a small minority arrived with both the click record and the buyer's own answer agreeing.
AI search was the worst case. The classifier logged its first-ever AI referral only in the final window, while buyers had been naming AI assistants all along.
Read on the dashboard alone, a channel that was already producing customers looked like it barely existed, which is exactly how a working channel gets defunded.
The solutionComparison pages, then a defensible number.
Three moves, in order. The first two create demand, the third makes it visible enough to fund.
BOFU focus
Own the shortlist queries
Best-of and alternatives pages against the queries buyers use to build a vendor list, shipped in batches rather than all at once so each cohort could be measured on its own performance instead of averaged into a single blur.
The full page set is now live. Bottom-funnel Google clicks rose 79% and sessions 48% in a window that included the summer slowdown, and the money term moved from page two to position 3.7.
Citations and brand mentions
Head-to-head pages as grounding sources
The comparison pages were written to be pulled into an AI answer, not just clicked. Within weeks they were surfacing at position 1 to 2 on long conversational queries, the evaluative phrasing buyers use with an assistant rather than a search box.
WaitWell now holds presence in roughly three of every four tracked AI answers, and its own pages are the cited source on a growing share of them.
Report against demand
Two signals, or it does not count
Nearly half the demo pipeline was filed as Direct, the bucket that means the classifier does not know. Every contact is now read on both the click record and the buyer’s own answer, resolved by a five-case matrix: agreement confirms, self-report recovers, conflicts resolve by channel type.
Demand-creation channels take the buyer answer, capture channels keep the click. Contacts with neither stay in a named Dark bucket.
That is how a channel the dashboard valued at almost nothing turned out to be running 2.9× the Realistic forecast.
The resultsAhead of forecast, on a number that survives scrutiny.
The programme was forecast before it was built, so it can be graded against its own promise rather than against a flattering comparison chosen afterwards. Four measures decide it: performance against forecast, whether the demand can be sourced at all, what the pages are doing, and whether the AI layer is moving.
Demos From Organic And AI Search Versus Forecast
Both scenarios were published before a single page shipped, so the programme is graded against its own promise. Cumulative demos from organic and AI search cleared the entire four-month Realistic forecast inside month one, and finished the period at 2.9× Realistic and 7.9× Conservative.
Customers tracked the same shape at 3.3× and 8.9×.
How this was derived · The Two-Witness Rule
The problem. Analytics has exactly one witness to where a lead came from, the referring URL. AI assistants strip that referrer, so a buyer who spent a week comparing vendors inside an assistant arrives looking like they typed the domain from memory.
The CRM files them as Direct, and Direct is not a channel. It is the system saying it does not know. Every quarter that goes unresolved, the channel doing the work looks like the channel doing nothing, and gets defunded.
The fix. We add a second, independent witness and rule on the pair rather than trusting either alone.
Neither witness is the truth on its own. One cannot see AI at all, the other remembers imperfectly. The rule is what turns two unreliable signals into one defensible number, and it produces the three tiers in the chart below:
Two things keep this honest on this account. Nothing is projected: the question sits on the demo form and every booker answered it, so the corrected number equals the observed number. And contacts that arrive outside the demo form carry no second witness, so they stay in the Dark bucket rather than being modelled into a channel that flatters the report.
Where The Demo Pipeline Was Filed
42% of all demo bookers were filed as Direct Traffic, the bucket that means the classifier does not know. Only 15% arrived with both signals already agreeing.
Read on the dashboard alone, nearly half of this pipeline had no source at all. Read on two signals, every one of them was placed, and nothing was projected: the demo form carried a 100% response rate, so the corrected number equals the observed number.
Three Reads Of The Same AI Buyers
The classifier found a fraction of the AI-sourced buyers. The buyers named AI 4× more often than the dashboard did, and reading both signals together surfaced 5×.
The missing ones were not lost, they were misfiled: they sit spread evenly across five different channel buckets, because a stripped referrer lands in Direct and a buyer who hears the name from an assistant then clicks a brand ad gets filed under the capture step. Quality held up under recovery: 9 in 10 booked a demo, 7 in 10 reached MQL or beyond, and the cohort produced the engagement’s first AI-attributed customer.
Program pages versus the site around them · 30 days over 30
The pages we built grew six to seven times faster than the site around them, in a window that included the summer slowdown, and the highest-value commercial term in the category moved from the middle of page two to position 3.7. Demo capture paused in that same window while visibility accelerated, which is the standard four to eight week lag between a page ranking and the meeting it books.
Competitive Presence In AI Answers
Share of tracked AI answers naming each brand in the category. The client is the only brand in the set gaining, up 15 points in a single window while two of the three competitors fall away.
Solid lines are measured; the shaded band carries each current trend forward, and on that trajectory the client passes the category leader in the next window.
The lesson is not that the program worked, it is that nobody could have proven it from the dashboard. Nearly half the demo pipeline was sitting in a bucket labelled Direct, and the fastest-growing channel looked like a rounding error until it was read on two signals instead of one.
Next step
Your GEO plan, built on your numbers.
Give us your domain and your pipeline goal. We map the prompts your buyers are asking, build your GEO marketing plan against them, and forecast the pipeline the next 90 days can realistically produce.
What we walk through