Google stopped distributing RuntimeWire. We kept growing.

Why it matters

A young AI-native publisher can lose discovery across an entire domain without a manual action or direct appeal path. That gives Google's automated systems broad power over which original reporting reaches readers.

Google Search impressions for RuntimeWire fell roughly 97% in five days. The English reporting did not suddenly change. Readers kept arriving. Our investigations kept spreading. Google went missing.

Today, The Wall Street Journal published a column written with AI under the byline of billionaire Stanley Druckenmiller. The page carried no visible AI disclosure when RuntimeWire reviewed it. Google found it immediately.

Fine. AI-assisted writing is allowed under Google's own rules.

We would like those rules applied to us, too.

What happened to RuntimeWire

RuntimeWire launched in May as an AI-native newsroom covering startups and technology. We use AI throughout the reporting process. Humans make the editorial decisions, review the sources, verify the claims, contact the people involved and take responsibility for every published word.

By August, Google Search was sending RuntimeWire tens of thousands of impressions per day. Search impressions reached 42,237 on August 12th.

Five days later, they were 1,327.

The collapse affected unrelated English stories across the site. Articles that had ranked normally stopped receiving meaningful impressions. New investigations struggled to appear even when RuntimeWire created the underlying search demand.

Our investigation into Anthropic's unreleased Project Parka is the cleanest example. RuntimeWire uncovered the project by examining artifacts inside Claude Desktop builds. Other publications and social accounts repeated our findings. Google's AI Mode could explain Project Parka using those details while initially citing downstream sources.

When asked for the original source, Google identified RuntimeWire and linked our investigation. When asked why RuntimeWire was absent from its first response, it apologized and said we deserved credit for breaking the story.

Google knew where the reporting came from. Its products used the information. Search still declined to surface the source.

The translation rollout

The collapse followed a large rollout of translated versions of existing RuntimeWire articles.

Those pages were created for international readers. They were not keyword variations, doorway pages or fabricated articles. They were translations of work we had already reported and published in English.

Google's systems apparently saw something else.

Search Console's "Discovered - currently not indexed" inventory surged to approximately 1,786 URLs. We responded aggressively. The translated pages were removed from Google, the sitemaps were rebuilt around the English publication, and the questionable inventory was retired.

The affected category subsequently fell from 1,786 URLs to 68. RuntimeWire's indexed share increased from approximately 52% to 73%.

The English reporting remained suppressed.

We cannot inspect Google's internal systems, so we cannot prove which classifier acted or whether the translation rollout caused the collapse. We can document the sequence:

• The English articles ranked.
• The translations appeared.
• Search distribution collapsed across the domain.
• The translations were removed from Search.
• Google's coverage reports recognized the cleanup.
• Distribution of the unchanged English articles remained near the floor.

Google provides no meaningful appeal process for an algorithmic sitewide judgment. We contacted people at Google. We publicly asked Google's Search Liaison for help. So far, we have received no explanation.

Then the Journal published this

On August 24th, billionaire investor Stanley Druckenmiller published "Let the Bond Market Speak" in The Wall Street Journal.

The following morning, Jeff Stein reported for NOTUS that Druckenmiller had used AI to write it.

"Of course I used AI," Druckenmiller said.

"I write everything using AI now for the same reason I use a calculator when I do math problems."

Druckenmiller disputed the suggestion that AI wrote the entire column and said he rejected many of its recommendations. His position was straightforward: the ideas were his, he approved the final copy, and his name was attached to it.

That sounds familiar.

In a statement, Journal opinion editor Paul Gigot defended Druckenmiller's use of AI and said the section's standard turns on authorship, standing and credibility rather than whether a contributor used the technology:

"AI is a fact of modern life. People will use it to assist in their work and their writing, including with research, checking grammar, editing and more. The question for us is whether what we publish from contributors reflects an author's original argument, and if the author has the standing and credibility to make it. In Stan Druckenmiller's case, we have had a relationship with him for many years, and nobody can doubt that his op-ed is his genuine opinion."

That is also a reasonable standard. It is the one RuntimeWire asks Google to apply consistently.

The Journal article was indexed immediately and amplified across Google, Reuters, Axios, the Financial Times, the Guardian and other publications. Its institutional history and Druckenmiller's expertise plainly carry enormous authority.

RuntimeWire publishes under a different set of practical conditions. We disclose our use of AI. We link the underlying records. We show readers how conclusions were reached. Our investigations frequently contain several independent sources and substantially more reporting than competing coverage.

We even tested that claim.

For a same-story reporting audit, RuntimeWire collected 1,996 articles published by us and 11 competing technology newsrooms. Candidate matches were clustered, publication branding was removed and 120 clean-text article pairs were used for the headline comparison.

RuntimeWire won 98 comparisons, tied one and lost 21, an 81.7% win rate. We published the methodology, the workbook and the raw files. We also reported the result that made us look worse: VentureBeat beat RuntimeWire 17 to four in its direct comparison, with one tie.

Google does not have to accept our audit as a ranking signal. Anyone can download the evidence and challenge it.

The claim that RuntimeWire is flooding the web with thin rewrites does not survive contact with the work.

Readers did not leave

Google Search distribution collapsed. RuntimeWire continued growing through direct readership, referrals, social distribution and people sharing stories with each other.

Our leading stories in the current public record include:

• An investigation into OpenAI agents rebuilding a secret message board: 14,761 views.
• The Project Parka investigation: 11,333 views.
• Reporting on an Apple engineer who says he was fired after refusing to send customer device identifiers to AT&T: 9,426 views.
• Coverage of Anthropic's developer backlash: 7,300 views.
• Our Grok Bot settings investigation: 5,927 views.

These are people choosing to read RuntimeWire while Google sends almost nobody.

We publish those figures in public, along with weekly results, methodology, corrections and the publications that cite our reporting. The numbers come from first-party logs with obvious automated traffic removed.

The newsroom has kept moving. The audience has kept growing. Sources keep bringing us material. Competitors keep following stories we break.

Google, pick up the damn phone

Google says generative AI content is eligible for Search when it is useful, original and created for people. Its spam policy targets scaled material created to manipulate rankings, regardless of whether humans or machines produced it.

We agree with that standard.

Our request is narrow: examine what happened to RuntimeWire and judge the English reporting on its actual merits.

If an automated system mistook a translation rollout for a sitewide spam operation, fix the false positive. If another problem exists, tell us what it is. If Google's real policy gives established publishers broad freedom to use AI while treating the same technology as a risk signal on young domains, say that plainly.

We have already removed the translations from Search. We have supplied the numbers. We have shown that Google's own AI systems recognize RuntimeWire as the original source of reporting they fail to cite. We have contacted Google privately and publicly.

The work continues either way.

RuntimeWire will keep breaking stories, publishing the evidence and earning readers one at a time. Search traffic would help. Permission is unnecessary.

This is Day 1.

Buckle up.

Google, pick up the damn phone.