Venture Step E087: Google Turned Off num=100

Google Turned Off num=100: What Venture Step E087 Saw

In Episode 87, Dalton Anderson treated Google's September 2025 num=100 change as more than a technical inconvenience. A small, unsupported search parameter had become infrastructure for tools, reporting, and automated collection. When the behavior changed, it exposed who depended on Google and how quickly a measurement shock could be mistaken for a market change.

The episode's durable insight is the dependency. Its claims about Google's motive, universal cost, bot traffic, AI competition, and scraping law require narrower language than the original monologue used.

The parameter that quietly became infrastructure

The num=100 parameter asked a Google results page to return a deeper set of links in one response. It was not a formally supported public product, but automated systems could use it to collect positions more efficiently.

Reports of intermittent failure appeared around September 10, 2025. Search Engine Land later published Google's confirmation that the results-per-page parameter was not formally supported.

That wording matters. Google confirmed a support boundary. It did not publish an explanation that the change was designed to block AI competitors, protect training data, increase revenue, or punish SEO tools.

timeline
    title E087 evidence timeline
    September 2025 : Reports that num=100 no longer works reliably
    September 2025 : Google says the parameter was not formally supported
    September 2025 : Ahrefs documents reduced deep-result coverage
    October 2025 : Ahrefs reports staged restoration
    November 2025 : Ahrefs reports broader top-100 coverage
    July 2026 : Venture Step separates confirmed change from inferred motive

Dalton's argument

Dalton described the change as Google "turning off the tap." In his reading, competitors and SEO systems had been using Google's result pages as an input to their own products. Making deep collection harder looked like a strategic move inside the AI arms race.

He also connected the change to implausible website-traffic numbers. If automated systems loaded results and visited pages, fewer automated requests could make analytics look more human. From there, he argued that search and authority measurement would need to change as answer systems became more important.

The episode moved from a specific interface change to a broad thesis about platform power. Google owns a crucial discovery surface. A dependent business can build on a behavior for years and still lack a promise that the behavior will continue.

That thesis survives the factual corrections.

What the vendor record establishes

Ahrefs published a first-party SERP-monitoring update on September 18, 2025. The company described reduced access to deep result pages, restored enterprise coverage in October, and a broader return to top-100 requests in November.

This record supports the episode's central operational point. At least one major vendor had a real dependency and had to adapt.

It also complicates the most dramatic version of the story. The old parameter behavior changed, but the vendor later restored coverage through its own system. Google did not permanently eliminate every path to deeper observation.

Semrush later examined a possible link between the parameter change and AI citation patterns. Its three-month study treated the explanation as a theory and compared actual movements across several systems. Correlation did not become motive simply because the dates aligned.

Where the episode overstated the evidence

The transcript described scraping public data as generally not illegal and primarily a terms issue. That is too broad for public guidance. Authorization, contracts, access controls, copyright, privacy, computer-misuse rules, jurisdiction, and conduct can matter. The E087 public package does not give a scraping-law conclusion.

The transcript also used a tenfold-cost claim. Ten smaller requests can replace one 100-result request in a simple example, but request count is not total cost. Vendors can alter depth, frequency, caching, providers, retry rules, and architecture. The effect is specific to each system.

The statement that analytics would become less inflated is a hypothesis to test. Google's current Search Console data documentation explains that data is processed and can differ from Analytics. Google Analytics says it excludes known bots and spiders to the extent possible. Neither source makes remaining traffic automatically human.

Why measurement changed even when the site did not

A rank tracker observes a search result through its own collection process. Search Console observes performance inside Google Search. Analytics observes configured behavior after a client reaches the site. Server records observe requests. Business systems observe outcomes.

If a third-party collector loses depth, its visibility chart can fall while Search Console clicks and qualified conversions hold. If automated result loads had generated impressions, Search Console totals may also shift even while human behavior holds. If an analytics tag fails, sessions can fall while server arrivals remain.

The E087 lesson is therefore diagnostic: a measurement change is not automatically a market change.

[[What Changed When Google Stopped Supporting num 100]] explains the technical layers without the episode's speculation. [[How to Diagnose a Sudden Search Analytics Change]] provides a source-by-source response. [[Bot Traffic vs Human Audience - What Analytics Can Tell You]] explains why bot classification remains uncertain.

The larger platform question

Dalton's most interesting question came near the end: how will publishers know whether they are an authoritative source when discovery systems increasingly answer questions directly?

That question belongs partly to Episode 102's answer-first and GEO work. E087 contributes a narrower warning. Authority measures will remain fragile if they depend on one provider, one parameter, one rank tracker, or one opaque score.

[[How to Build SEO Measurement Without One Search Parameter]] proposes a more durable system. It starts with business decisions, gives each source one job, records limits beside metrics, and preserves independent evidence where failure matters.

Episode 87 did not prove that Google declared war on AI competitors. It documented the feeling of a platform boundary suddenly becoming visible. The practical response is not to guess the platform's private motive. It is to identify the dependency, preserve the evidence, and build measurement that can survive the next interface change.

The source transcript remains preserved without alteration. This page distinguishes Dalton Anderson's September 2025 interpretation from sources reviewed on July 27, 2026. AI assistance was used for research organization, drafting, and validation. Publication remains unauthorized.