Almost nobody buying a contact database asks where the records came from. They ask about match rates, coverage in their territory, refresh frequency and price per credit. Those are the right questions for a market where supply is stable. Supply stopped being stable this summer, and the questions have not caught up.
This is not a compliance lecture. It is a continuity problem wearing a legal costume. If a supplier three layers upstream of your sequencing tool loses access to a source, your reply rates fall in a way that looks like copy fatigue, and your team will spend a month rewriting emails that were never the issue.
Two court cases that changed the B2B prospecting data question
In the Northern District of California, before Chief Judge Yvonne Gonzalez Rogers, Google is suing SerpApi. A July ruling knocked out the broad theory that scraping results is itself a copyright violation, which was widely read as a defeat for Google. It was, and Google came back narrower. The amended complaint filed on August 10 argues that its licensors authorized the technical access controls, which would make circumventing them a violation under Section 1201 of the Copyright Act regardless of what the scraping itself is.
The evidence Google supplied is the interesting part. It points to licensing agreements that direct Google not to let third parties extract and independently commercialize the licensed content, and says one licensor specifically asked for technical measures when it suspected unauthorized scraping. That licensor is Reddit. Google also cites its own privacy policy as authorization from the individual users whose photos and reviews appear in results. SerpApi's public response on August 11 was that it remains confident in the strength of its position, and the filing has been reported in detail.
Meanwhile in New York, Judge Paul Engelmayer allowed Reddit's own claims against SerpApi and Perplexity to proceed in late July. Reddit is in a stronger position there because it owns or licenses the content directly from its users rather than through third parties. Two courts, two theories, one supply chain.
“We remain confident in the strength of SerpApi's position.”
None of this is about cold email, and that is exactly why it matters. The legal argument being tested is that a contract between a platform and its licensors can convert an ordinary anti-bot measure into an access control with statutory teeth, and that the obligation travels downstream to anyone commercializing the extracted content. If that argument lands, it does not stay inside the search tool category. It describes the mechanics of every enrichment provider that reads a platform which has both anti-bot measures and licensing deals, which by 2026 is all of the large ones.
The four supply chains your data actually comes from
Every B2B prospecting data product is one of four things, or a blend that the vendor would rather describe as proprietary. Knowing which one you bought tells you almost everything about how it will behave under pressure.
| MODEL | HOW RECORDS ARE OBTAINED | CONTINUITY RISK | WHAT YOU SHOULD PAY FOR IT |
|---|---|---|---|
| Licensed | Paid agreements with the source platform or publisher, with terms on downstream use | Low, but the licensor can change terms at renewal and often does | The most, and it is usually worth it for the core list |
| Contributed or co-op | Users of the vendor's tools share their own address books and CRM records back into a pool | Low legally, variable in quality, and coverage skews to whoever the contributors sell to | Mid range, with heavy verification of your own segment |
| Public and scraped | Automated collection from sites and platforms without a license | Highest, and the thing both live cases are about | Least, and never as the only source under a revenue forecast |
| First party and observed | Your own CRM, site, product and event data, plus signals you generate | None, and it is the only asset nobody can revoke | Whatever the engineering costs, because it compounds |
Most commercial datasets are a blend, and blends are where the risk hides. A vendor can accurately describe a product as licensed when the licensed portion is company firmographics and the contact records underneath are scraped. Ask about the blend by field, not by product. Coverage of job titles, direct dial numbers, personal email addresses and intent signals often come from four different places inside the same subscription.
Illustrative allocation of an outbound data budget under a continuity-first policy, versus the pattern we typically find on first review. Illustrative, not measured.
The vetting questions that get real answers
Vendor questionnaires produce vendor answers. These five questions produce information, because each one is hard to answer vaguely without the vagueness itself being the answer.
What breaks first when a supplier loses access
The failure is never announced and it never looks like a data problem from inside the sequencing tool. It looks like performance decay, and performance decay gets assigned to the copywriter.
The instrument that catches all four early is boring and almost nobody runs it. Hold out a fixed sample of 500 accounts, re-enrich them monthly with the same vendor, and track completeness by field over time. It costs a rounding error in credits and it converts an invisible decay into a dated line on a chart. When the line steps down, you know it is supply, and you know when.
Writing B2B prospecting data terms into the contract
Procurement will happily add clauses if someone tells them which ones. Four are worth the negotiation.
First, a source change notification clause with a number of days in it. Second, a provenance warranty covering the specific fields you depend on, rather than a general statement of lawful collection. Third, an audit right on a sample of records, exercised annually, because the right to ask is what keeps the answer honest. Fourth, an exit provision that lets you take the records you have already paid for out in a usable form, with the enrichment history attached.
Expect the notification clause to be the one that gets pushed back on hardest, and hold the line on it, because it is the only one of the four that helps you before the damage rather than after. Thirty days is a reasonable ask. Fifteen is achievable at renewal if you are willing to trade something on term length. What you are buying is the chance to run a parallel source for a month rather than discovering the change through a bounce report, and that difference is worth more than most of the discount your procurement team will spend the same meeting chasing.
Then reduce the dependency itself, because contract terms are compensation for a risk rather than a removal of it. Every outbound program should be moving budget toward data nobody can revoke: your own CRM history, site and product behavior, event and community participation, and the accounts that engaged and did not convert last year. That set is smaller than a purchased database and converts several times better, and its supply chain is a conversation with your own engineering team. It is the first thing we rebuild in a cold email engagement, and in most B2B programs it turns out that the highest performing segment was already sitting in the CRM being ignored in favor of a fresher list.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Tyler leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.