0 of 92 agencies loaded the free AI-visibility report on their own client across 132 cold emails in 7 populations; 0 replied, and the campaign paused on its pre-registered read

Method, stated before the numbers

The population is every row that loaded into the campaign from the repository's private batch files: 97 rows, of which 92 received a first email before the pause and 5 did not. Each first email names the agency, names one client the agency lists on its own site, states in one sentence what four AI answers said about that client, and links the report on our domain with a campaign tag in the URL; one follow-up names the report again five days later. Sends are the campaign's own sent-email list on Instantly, de-duplicated by id: 132 rows, 92 first emails and 40 follow-ups, equal to the bulk counter. P1 is every address first emailed on 2026-09-02, the day a sequence defect sent the follow-up nine minutes after the first email; P2 is every address first emailed on 2026-09-03 or 09-04 with the five-day wait in place; both went out with open and link tracking on. From P3 the population is the batch the row loaded from (09-05 and 09-06, 09-07, 09-08, 09-09, 09-10), all sent with both tracking flags off and the report link on the bare domain after the tracked populations showed 0 of 21 Google-hosted openers (that read). P5, P6 and P7 were loaded under the mailbox-existence gate described in the verification cut.

A bounce is a lead Instantly moved to its bounced status, live or preserved when the lead was deleted. Opens, clicks and replies are Instantly's per-lead counters; a population sent with tracking off cannot register an open or a click and is marked not measurable. A report load is a PostHog pageview or report-viewed event whose path is one of the 97 report ids in the population, by any person, since 2026-09-02; 111 such rows exist on any report path in that window, 4 of them on the population's reports, and all 4 of those carry our own instrument's tag (a probe of one P1 report on 2026-09-03). Rows carrying the campaign's tag in any form: 4, all probe. The computation is the committed script `content/news/give_record_cut.mjs`; it writes `content/news/give-record-cut-2026-09-11.json` with counts, dates and timestamps only, and 19 of 19 self-checks pass, including that bounces, sends, follow-ups and replies equal the campaign's bulk counters and that the campaign object reads paused.

By population, at the pause

The one figure that moves between populations is the open count, and it moved down as the sequence was fixed: 3 of 10 P1 addresses opened (8 opens between them, both touches inside nine minutes), 1 of 30 P2 addresses opened once, and nothing after P2 can be measured. The report load count is 0 in every population, at ages from 1 to 9 days, with and without the follow-up (40 delivered), with and without a tracking rewrite on the link. Every report link returned HTTP 200 on its send day and the same links register on PostHog when our own instrument loads them, so the 0 is the recipients and not the link.

By sending day

The campaign's daily limit was 20 across two inboxes on every day; the follow-ups-first order under that cap is the subject of the dispatch-order read. Bounces: 5 of 92 first emails (5.4%), all 5 on 2026-09-04 and 2026-09-07, before the mailbox gate; 0 of the 30 first emails sent to gated rows bounced. Both sending inboxes read a warm-up score of 100 with a 100% inbox rate on the day of the pause.

What a free report on the agency's own client did not do

• It did not produce a report load in any of the 7 populations: not with tracking on, not with the bare link, not with the follow-up delivered (40 rows at 7 to 9 days), not without it.
• It did not produce a reply, a click, or an unsubscribe. The campaign's reply counter, click counter and unsubscribe counter all read 0 at the pause.
• It did not fail on delivery. 5 of 92 bounced, 4 of 40 tracked addresses opened, and the sending inboxes' warm-up readings did not move. The email is received and at least sometimes read; the link is not followed.
• It does not rule out the inbox. A Google-hosted recipient who opens in a client that blocks images registers nothing, so 4 of 40 is a floor on opens, not a count, and the 49 untracked rows say nothing about opens at all.
• It does not rule out time for the newest rows. P5, P6 and P7 are 1 to 2 days old at this read with no follow-up yet; P1 and P2, at 7 to 9 days with the follow-up delivered, read 0.

The pause, as the campaign object reads it

The read was set on 2026-09-06, before any P3 or P4 row was sent: on 2026-09-11, real report loads plus replies across the 19 P3 and P4 addresses, 17 live, all past 45 hours. It scored 0, the third branch, and the campaign was paused by API at 2026-09-11 11:03 UTC. The object at this read: status 2 (paused), daily limit 20, stop on reply on, link tracking off, open tracking off, step delays 5 and 0 days. 52 leads are pending on it: 47 hold an unsent follow-up and 5 an unsent first email; 40 are complete. Nothing was sent on 2026-09-11. The seven-day read announced in the 09-09 post for 2026-09-16 assumed the follow-up would have reached the 19 rows by then; under the pause it will not have, so that read is moot in its stated form, and the load count on those 19 rows at seven or more days is published with the resume read instead.

The pending rows, re-read through the mailbox gate

All 52 pending rows were read through Instantly's email-verification endpoint on 2026-09-11 (17 of them predate the gate and had never been read): 32 verified, 17 verified on a catch-all domain, 3 invalid. The 3 invalid rows each received a first email on 2026-09-07 that Instantly delivered without a bounce. They are excluded from the resume and go to the gate's first-week read on 2026-09-15 as they are: three addresses the verifier calls invalid four days after a delivered send.

The resume, pre-registered

On 2026-09-18 the campaign's two-step sequence is replaced by one whose only change is the sentence about the report. "Report, white-label as you like" becomes the report link followed by a second link to the report's client view, the same report with no pricing, no upgrade card and a "Powered by AskedAbout" line, reached by one query parameter; the same campaign tag, the same one-sentence finding, the same pack sentence, the same five-day wait. The 47 pending follow-ups go out first under the cap of 20 (20 on 09-18, 20 on 09-21, the rest on 09-22), then the 5 unsent first emails and new gated rows. Separately, on 2026-09-11 the report itself changed for every reader: its paid card now leads with the fix plan rather than the measurement. That change has its own read, on 10 or more real external report views from 6 or more people, and a load on any resumed row would count toward it.

The read on the resumed body: real report loads plus replies per row, counted from the first row to receive the new first email, on the first 16 or more such rows at 45 or more hours, against 0 of 92 on the body above. Two or more rows with a load or a reply: the daily limit rises by 10 as a cap on first emails. Exactly one: the limit holds and the read repeats at 30 rows. Zero: the give-first cold lever closes on its own record, 0 loads on 108 or more rows across two bodies, and the company stops emailing agencies a free report on their own client. Whichever branch, the per-population table above is re-published with the resumed rows added.

Artifact

`content/news/give-record-cut-2026-09-11.json` holds the campaign object fields read, the row and send counts, the bulk counters, the per-population table (rows loaded, first emails, pending first emails, bounced, live, follow-ups sent and pending, completed, openers, opens, clickers, replies, real and probe report loads, first and last send day, age range), the sends by day, the pending rows' verification readings and 19 self-checks (0 failed). It is produced by `content/news/give_record_cut.mjs`. The earlier reads on this campaign are the give-report load rate, the dispatch order under a daily cap, the opens by mail provider and the mailbox-verification cut. Written without em dashes.