Learn
Case Study · July to November 2026

First client,
documented.

A dual credential broker associate and state certified real estate appraiser serving Northern New Jersey. Published anonymously at the client's request; every figure below is unedited.

The Headline Receipt · Third-Party Measured
251
Up from 26 three months earlier
Copilot citations in Microsoft's 3-month AI Performance window

The window ending August 8 held 26 citations, all beginning around July 15 when the content system launched. The window ending November 5 holds 251. Same measurement, same source, three more months of publishing. Average cited pages is still 1: one page is doing this work.

Microsoft-sampled data from Bing Webmaster Tools · 3-month windows verified August 8 and November 5, 2026
69.09%
Up from 30.95% three months earlier
Citation share on the query "accurate Hoboken home valuation methods"

Microsoft's topic: Home Appraisal & Valuation. In plain terms: the site went from being cited in roughly 3 of every 10 sampled answers for that query to roughly 7 of every 10. That one query accounts for 76 of the 251 citations; the rest are spread across queries Microsoft does not break out individually.

Microsoft-sampled data from Bing Webmaster Tools, verified November 5, 2026
The Receipts · Straight From The Source
Bing Webmaster Tools AI Performance overview for the client site, with the domain redacted: 251 total Copilot citations in the 3-month window ending November 5, 2026, with the trend rising through September and October

The AI Performance overview today: 251 total Copilot citations in Microsoft's 3-month window, with the daily trend building through September and climbing through October.

Captured November 5, 2026 · Bing Webmaster Tools · Microsoft-sampled data
Bing Webmaster Tools grounding query table with the client domain redacted: accurate Hoboken home valuation methods, Research intent, Home Appraisal and Valuation topic, 76 citations, 69.09 percent citation share

The grounding query itself: "accurate Hoboken home valuation methods," filed by Microsoft under Home Appraisal & Valuation with a Research intent, now at a 69.09% citation share across 76 sampled citations.

Captured November 5, 2026 · Bing Webmaster Tools · Microsoft-sampled data
The same two views, three months earlier
Bing Webmaster Tools AI Performance overview with the domain redacted, captured August 8, 2026: 26 total Copilot citations in the 3-month window, activity beginning mid-July

26 total citations, all beginning mid-July.

Captured August 8, 2026
Bing Webmaster Tools grounding query table with the domain redacted, captured August 8, 2026: the Hoboken home valuation query at 26 citations and a 30.95% citation share

The same query at a 30.95% citation share.

Captured August 8, 2026
The Speed Story
12 days
First AI citations observed 12 days after the first posts published

First posts went live in early July. First citations were logged July 14. First observed, not caused: we log what the engines do and we date every log entry. Faster than we tell clients to expect. Our own guidance says citations typically follow 4 to 8 weeks of held rankings, and we stand by that. This one moved early.

July
The July baseline: 12 standardized queries, restated under a strict rule

We run the same 12 queries every month. Under our strict definition (cited means the domain appears in the AI answer's own sources; named means named in the answer text without a link), July's baseline reads: ChatGPT 0 of 12, Perplexity 1 cited, Google AI 1 named.

  • "Hoboken condo market 2026": the report cited in Perplexity's sources panel. The genuine article.
  • Named first in a Google AI answer for "real estate appraiser who also sells homes in New Jersey", with no link back. Named, not cited: the strict rubric counts them differently.
Month Two · The Trend

Methodology note, August 2026: we tightened the definition of a citation to in-answer sources only and restated July under the same rule. One July entry turned out to be organic placement near an AI answer, not a citation inside it, and it no longer counts. We would rather restate a month than inflate one. The full standard is published at how we measure AI visibility.

The monthly panel, month over month
Engine July (restated) August
ChatGPT 0 of 12 Named first on 2 of 12: both dual credential queries
Perplexity Cited on 1 of 12 0 of 12 visible; the recheck found the site still in the source list, below the fold
Google AI Named on 1 of 12 0 of 12
Manual panel run August 4, 2026 · Filtrs citation log
+2
Gained in August

ChatGPT now names the client first on both dual credential queries: "real estate appraiser who also sells homes in New Jersey" and "NJ real estate agent who is a certified appraiser". The appraiser-broker moat, answered by name.

−2
Lost in August, reported anyway

Perplexity's July citation on the Hoboken condo report slipped below the visible fold: the August 8 recheck found the site still in the source list at position 6, outranked by fresher-dated monthly reports. A demotion, not a disappearance, and the fix is freshness. And Google AI now answers the appraiser query with a general explainer that names no individuals at all.

The Frequency Layer

What daily sampling sees that a single run cannot.

Since August 5 we also sample ChatGPT by API, three samples per query, every day. The same questions, asked repeatedly, tell a deeper story than one screenshot:

  • "How much is my house worth in Bergen County NJ": the Bergen valuation page cited in 3 of 3 samples, and holding the position 1 citation in the August 7 sweep. A stable citation on the money query, invisible to the single manual run.
  • "Hoboken condo market 2026": the report cited with the client's numbers quoted inline, in 1 of 3 samples and again in the August 7 sweep. The July Perplexity demotion, partly offset on a different engine.
  • Both dual credential queries: named first in most samples (2 of 3 and 3 of 3), and the appraiser-who-sells query upgraded to a full citation in the August 7 sweep: the site in the source list and the client's business profile at position 1.

One screenshot says visible or invisible. Frequency says how often. This is the reason the panel runs repeatedly instead of trusting any single answer.

Automated ChatGPT sampling, daily since August 5, 2026, 3 samples per query · Filtrs citation log
Meanwhile, on classic Google

"Hoboken condo market 2026": the report moved from organic position 4 in July to position 1 in August, with the client's statistics quoted in the result snippet.

Google Search, logged August 4, 2026
Three Months Later · The Search Trajectory
3,787
Google impressions in the first three months of measurement

From a domain that showed roughly 7 visible queries in its early weeks. Small in absolute terms, and we say so: this is a young domain's trajectory, not a traffic story. The direction is the finding.

Google Search Console, July 27 to October 27, 2026
8.6
Average Google position across those impressions

A four-month-old site averaging page one across everything it surfaces for. The content earns positions first; volume is the slower curve behind it.

Google Search Console, July 27 to October 27, 2026
4.4%
Click-through rate on the estate valuation guide

Roughly one in 23 searchers who saw the guide clicked it. High-intent pages written for one specific question earn clicks even while a domain is still young.

Google Search Console, July 27 to October 27, 2026
87 → 18
Average position on the seven hiring-intent queries

The queries someone types when they are choosing an appraiser moved from average position 87 to 18 over the window. Still not page one, and we report that too; the climb is the point.

Google Search Console, July 27 to October 27, 2026
The Measurement System, Upgraded

From manual panel to recorded instrument.

The July and August numbers above came from a 12-query panel run by hand. Since mid-August the checks run themselves: 25 tracked queries, asked on a fixed schedule across ChatGPT, Google AI, and Perplexity, with every answer and every cited source kept as append-only history. Engines are never pooled, and every count is published with its denominator.

  • November is the baseline month: the first complete month of automated citation data. Per our own standard, a panel without history is queries monitored, not a benchmark, so November's job is to set the number: citation rate per engine across the 25 tracked queries.
  • December gets the first comparison: measured against November, engine by engine, published here. Any citation that existed in the prior month and is absent gets named, with the query and the engine.
Automated citation sweeps since mid-August 2026 · Filtrs citation log
The Starting Condition

A genuinely cold start.

Brand-new domain. Zero search history. First content published early July 2026, with six data-backed articles live as of late July. In the early weeks, only about 7 distinct queries were even visible in Google Search Console. That is the "before" picture, and it is what makes a 12 day citation timeline the story.

Supporting Facts
#1 on Bing for the brand query

Fully indexed and served by Bing, ranking first for the client's own name and brand.

Bing Webmaster Tools, confirmed July 27, 2026
Position 1 on Google for "hoboken neighborhoods map"

The neighborhood guide with an interactive map outranks the portals for the query it was built for. The guide plus map format works.

Google Search Console, July 2026
What the results ride on

A verified market data pipeline with a fact checker that has refused drafts containing wrong numbers. Self-updating county data pages. A dynamic llms.txt. Speakable schema. IndexNow pings on every publish. This infrastructure is the service: every Filtrs client runs on the same machine.

Why this client is not named.

The client asked to stay anonymous, so the name and domain are redacted here and inside the screenshots. Nothing else is altered: the queries, counts, dates, sources, and losses are exactly as recorded. Our measurement standard explains how every figure on this page was collected.

What This Does and Does Not Prove

This demonstrates that the client became more visible and citable to AI search systems. It does not yet prove the visibility produced a client.

AI answers are volatile. Every result on this page is logged with its query, date, platform, and screenshot, and is reported as observed, not as a permanent ranking.

And for balance: August recorded losses alongside gains. Perplexity, which cited the site in July, kept it only below the fold in August. Google AI now answers the appraiser question with a general explainer that names no one. Losses are published here with the same prominence as wins. That is how the rest of this page stays believable.

Want receipts like these?