Which B2B Data Tools Google AI Names: 93 SERPs, 25 Questions, Aug 2026

On August 15, 2026 Cleanlist measured how Google answers the questions technical buyers actually type about B2B data: 93 usable search queries and 25 Google AI Mode buying questions. Google AI Mode names Apollo in 16 of the 25 answers (64%), Clay in 14 (56%), ZoomInfo in 14 (56%), Hunter in 8 (32%), and Cleanlist in 7 (28%). An AI Overview appeared on 84 of the 93 queries that returned a usable SERP (90.3%), and cleanlist.ai is cited in 11 of those 84, the ninth most-cited domain in the set. Both datasets are published below under CC BY 4.0.

Last updated: August 15, 2026. Every number on this page was collected on August 15, 2026 (UTC timestamps run into the early hours of August 16). Nothing here is carried over from an earlier study, and the tool-naming counts describe that single day only.

How many B2B power-user searches now show a Google AI Overview?

Almost all of them. Cleanlist issued 98 queries on August 15, 2026 covering MCP servers, agents, enrichment APIs, automation, lead sourcing, GTM engineering and waterfall enrichment. Five of those returned an upstream SERP error (DataForSEO status code 40101) and produced no result at all, so they say nothing either way about AI Overview presence and are treated here as non-response rather than as misses. Of the 93 queries that returned a usable SERP, 84 carried an AI Overview, which is 90.3%. Counting the 5 failures as misses instead gives 84 of 98, or 85.7%, and that is the floor. Both numbers are on this page because the published CSV carries exactly 93 SERP rows and 25 AI Mode rows, so 90.3% is the figure you can recompute from the open data and 85.7% is the most conservative reading of the same collection.

How many sources does a Google AI Overview cite on these questions?

Roughly nine. Across the 84 AI Overviews in this dataset the ai_overview_source_count column sums to 761 source slots, a mean of 9.06 sources per AI Overview. The spread is wide. Two queries pulled 24 sources each (lead generation mcp server and linkedin api alternative), lead generation api pulled 23, and b2b data api, company data api and b2b data mcp server pulled 20, 20 and 19. At the other end, zapier lead enrichment pulled 3 and net new leads pulled 3. Cleanlist appears in 11 of those 84 source sets. The gap between roughly nine sources read and one or two products named in the visible text is the whole subject of this report.

Which websites do Google AI Overviews cite on B2B data questions?

YouTube, by a wide margin, then Reddit. Neither is a B2B data vendor. Below are the domains cited by the most distinct AI Overviews in this set of 84.

Two of the top four are user-generated platforms. A video and a forum thread out-cite every vendor marketing site in the category, including Cleanlist's, which sits ninth at 11.

Which B2B data tools does Google AI Mode name in 2026?

Apollo, in almost two thirds of answers. Cleanlist put 25 power-user buying questions to Google AI Mode and counted which of 20 dictionary tool names appeared in the visible answer text. Sixteen of the 20 were detected at least once, plus Cleanlist, giving the 17 rows below. UpLead, Lemlist, Unify and Common Room were in the dictionary and were named zero times.

The distribution has a long tail. Three names clear half the question set, and eleven of the seventeen sit at 24% or below.

Where does Cleanlist rank in its own measurement?

Fifth of seventeen, at 28%. Google AI Mode named Cleanlist in 7 of the 25 power-user buying questions. That places Cleanlist behind Apollo (64%), Clay (56%), ZoomInfo (56%) and Hunter (32%), and ahead of Instantly, Lusha, People Data Labs, Seamless, Clearbit, Explorium, RocketReach, Artisan, Cognism, Smartlead, AiSDR and RB2B. It is a better result on this question set than the 9.1% share Cleanlist recorded across 712 answers and 51 tools in the broader AI Visibility Index published on August 7, 2026, and the reason is narrowness rather than strength: these 25 questions are heavily weighted toward enrichment and API topics that Cleanlist has published on for months.

Which questions does Google AI Mode answer with Cleanlist in the text?

Seven, and they cluster tightly. The answers naming Cleanlist are: "best lead enrichment api for developers in 2026", "what is the best people search api", "how do gtm engineers build prospect lists in 2026", "what is waterfall enrichment and which tools do it", "which enrichment provider has the best email find rate", "how much does b2b contact data cost per record in 2026" and "what is the cheapest lead enrichment api". Three are enrichment questions, three are API or pricing questions, one is about GTM engineers. Every one of them maps to a long-form page Cleanlist already publishes. None of them is about agents, automation or sourcing.

Which questions is Cleanlist completely absent from?

The agent surface, entirely. Cleanlist is named in 0 of the 4 MCP questions in this set: "is there an mcp server for b2b lead data", "what mcp servers can find and enrich leads", "how do i give claude access to b2b contact data" and "how do i build a prospecting agent with claude code". The SERP half of the dataset says the same thing from the other direction. Fourteen MCP queries were measured (mcp server, best mcp servers, mcp server list, sales mcp server, mcp server for sales, b2b data mcp server, crm mcp server, lead generation mcp server, claude mcp server, mcp server claude desktop, how to use mcp servers, what is mcp server, model context protocol, claude for sales). Thirteen returned an AI Overview, mcp server claude desktop did not, and cleanlist.ai is cited in none of the fourteen. Cleanlist is also absent from every automation question and every net-new-sourcing question in the set.

Which AI Overviews cite cleanlist.ai?

Eleven, and all eleven sit in three topics. The queries whose AI Overview cites cleanlist.ai are lead generation api, person enrichment api, proxycurl alternative, people data labs alternative, clearbit api alternative, gtm engineer, what is a gtm engineer, waterfall enrichment, waterfall enrichment tools, email waterfall and phone waterfall enrichment. Five are API or API-migration queries, four are waterfall enrichment queries, and two are GTM-engineer queries. The other 73 AI Overviews in the dataset, covering MCP, agents, automation, sourcing, lead databases and AI SDR tooling, read other people's pages. Cleanlist holds a top-25 organic position on 16 of the 93 SERPs measured, so the citation gap and the ranking gap are the same gap.

What is the difference between being cited and being named?

A citation is the engine reading your page. A name is the engine writing your brand into the sentence the buyer reads. They are separate outcomes produced by separate work, and this dataset records them as separate columns. In the published CSV, cleanlist_cited is set from the source URL list attached to the answer, and the naming figure is set from a word-boundary match on the visible answer text after every URL has been stripped out of it. Cleanlist tracks both because a buyer scanning a Google answer sees the names and rarely opens the sources. Being in the source set and being in the shortlist are not the same commercial event.

Why does confusing a citation with a name inflate the vendor running the study?

Because a brand name usually lives inside its own domain, so any regex that scans raw answer markdown will match a brand inside its own citation URL. Cleanlist made exactly this error. The AI Visibility Index first published on August 3, 2026 scanned Google AI Mode markdown with inline source URLs still in it, so every citation of a cleanlist.ai page scored as an answer naming Cleanlist. The correct figures for that study are 712 answers, 5 engines, 51 tools, Cleanlist at 9.1% share of voice and rank 17 of 51. The pre-fix numbers were withdrawn. The detector used for this August 15 dataset strips URLs first:

// A citation is the engine READING us. A name is the engine SAYING us. Strip the
// source URLs before matching or every citation counts as a name. This is the exact
// bug that produced the retracted figure on Aug 3.
const stripUrls = (t = '') => String(t)
  .replace(/https?:\/\/\S+/g, ' ')
  .replace(/\[[^\]]*\]\([^)]*\)/g, ' ')
  .replace(/\b[a-z0-9-]+\.(ai|com|io|co|net|org)\b/gi, ' ');

Any vendor running an AI visibility dashboard should check this first. The bias is not random. It systematically inflates whichever brand publishes the pages the engine cites, which is disproportionately the brand running the measurement.

Did the ghost citation gap appear in this dataset?

Not on this surface, and the reversal is worth stating carefully. Across the 25 Google AI Mode questions, cleanlist.ai was cited in 6 answers and Cleanlist was named in 7, so naming exceeded citation by one. That is the opposite shape to the Ghost Citation Report Cleanlist published on August 10, 2026, which found 30 citations against 1 name across 58 Google AI answers. The difference is almost certainly the query set rather than a change in Google. The August 10 keywords were drawn mostly from keywords cleanlist.ai already holds an organic position on, 27 of the 37 from Cleanlist's own AI Overview citation gap, so that sample is anchored to Cleanlist's ranked footprint. These 25 are buying questions selected before any result was seen. Read together, the two say the gap is a property of the question, not a constant.

Which power-user searches have no AI Overview at all?

Nine of the 93 usable SERPs. They are mcp server claude desktop, ai sdr, vibe prospecting, email finder api, apollo api, lead enrichment automation, make.com lead generation, data enrichment pipeline and contact match rate. The pattern is loose but visible: short navigational or brand-anchored queries (apollo api, ai sdr) and very thin-volume queries tend not to trigger an overview, while explanatory and comparative queries almost always do. On those nine, a classic blue link is still the whole result, which makes them the cheapest ranking targets in the set for any vendor, Cleanlist included.

What does this mean for anyone trying to be visible in AI answers?

Three things this data supports, stated as findings rather than laws. First, an AI Overview is now the default on technical B2B data queries at 84 of the 93 that returned a usable SERP, so a page written only for a click position is being read by a system that may never produce the click. Second, citations and names have to be counted separately, because the same domain can be ninth most-cited in a set (Cleanlist, 11 of 84) while being named in a minority of answers. Third, presence is topic-shaped rather than domain-shaped: Cleanlist is named in 7 answers about enrichment, APIs and pricing and 0 answers about MCP, agents, automation or sourcing, from the same domain on the same day.

Method: how exactly was this measured?

Everything below is fixed and reproducible.

• Date: August 15, 2026. Collection timestamps run 2026-08-15 into 2026-08-16 UTC.
• Engines: Google organic SERP with AI Overview (v3/serp/google/organic/live/advanced) and Google AI Mode (v3/serp/google/ai_mode/live/advanced), both via the DataForSEO SERP API.
• Location, language, device: United States (location code 2840), English, desktop.
• Sample sizes: 98 SERP queries issued, 93 returned a usable SERP, 84 of those carried an AI Overview. 25 Google AI Mode buying questions, all 25 returned an answer.
• Tool dictionary: 20 competing tool names, case-insensitive substring match on the cleaned answer text. 16 were detected at least once.
• "Cited" means the domain appears in the answer's source set. For AI Overviews it is read from the citation objects; for AI Mode it is read from the link targets.
• "Named" means the brand appears in the visible answer text after inline source URLs, markdown link syntax and bare domain.tld tokens have been stripped. Cleanlist is matched with a word boundary (/\bcleanlist\b/i).
• Counting rule: a tool counts once per answer no matter how many times it appears.

Limitations: what can this dataset not tell you?

Six limits, stated so the numbers can be discounted correctly.

• One engine family, one country, one day, 25 questions. This is Google only, United States only, English only, desktop only, and a single collection run. It is a snapshot, not a trend, and it cannot be extrapolated to ChatGPT, Claude, Perplexity or a non-US buyer.
• A mention is not a recommendation. An answer saying "Apollo's phone coverage is thin, use something else" scores a mention for Apollo. Every share figure counts naming frequency and nothing else.
• Dictionary blindness. Only 20 tool names were searched for. A tool outside that list is invisible here regardless of how often Google names it.
• The URL stripper removes bare domains, so counts are a floor. Because \b[a-z0-9-]+\.(ai|com|io|co|net|org)\b is stripped before matching, a tool that Google referenced only as hunter.io or clay.com and never as a plain word is not counted. Every share figure in this report is a conservative floor.
• The two matchers are not identical. Competitor names use a case-insensitive substring test and Cleanlist uses a word-boundary test. Substring matching is the looser of the two, so whatever bias that asymmetry introduces runs in favour of the competitors' counts rather than Cleanlist's.
• Five queries failed. Five of the 98 SERP requests returned an upstream error and produced no data, which is why this page publishes both 84 of 98 and 84 of 93.

How can I reproduce or recompute this?

Both datasets are published under CC BY 4.0, so you can redistribute, remix and build on them commercially as long as you credit Cleanlist.

• Power-user AI visibility dataset: 118 rows, one per query. Columns are section (ai_mode_question or serp_keyword), key, has_ai_overview, ai_overview_source_count, cleanlist_cited and cleanlist_named_or_position. The last column carries a 0/1 naming flag on AI Mode rows and an organic position on SERP rows, so read it by section.
• AI Mode tool mentions dataset: 17 rows, one per tool, with answers_naming_it, of_25_questions and share_percent.

To rerun it, issue the same 98 keywords and 25 questions against the two DataForSEO endpoints listed in the method with location 2840, language en, device desktop, then apply the stripUrls function above before any brand matching. If your numbers differ from ours, publish them. A category with two independent measurements is in better shape than a category with one, including when the second contradicts the first.

FAQ

Which B2B data tool does Google AI Mode name most often?

Apollo. In Cleanlist's August 15, 2026 measurement of 25 power-user buying questions, Google AI Mode named Apollo in 16 answers (64%), Clay in 14 (56%), ZoomInfo in 14 (56%), Hunter in 8 (32%) and Cleanlist in 7 (28%). Sixteen of the 20 tool names in the dictionary were detected at least once; UpLead, Lemlist, Unify and Common Room were named zero times. These figures count how often Google wrote a name into the visible answer on one day. They do not rank the products, do not measure accuracy or price, and a critical mention counts the same as a recommendation.

What percentage of B2B data searches have an AI Overview?

90.3% in this sample. Cleanlist issued 98 power-user B2B data queries on August 15, 2026 covering MCP, agents, enrichment APIs, automation, sourcing, GTM engineering and waterfall enrichment. Five returned an upstream SERP error and produced no data at all, so of the 93 that returned a usable result, 84 carried an AI Overview, which is 90.3%. Counting the 5 failures as misses gives the conservative floor of 84 of 98, or 85.7%. Nine usable SERPs had no AI Overview at all, including apollo api, ai sdr, email finder api and vibe prospecting. The full per-query breakdown is in the open CSV.

Which websites do Google AI Overviews cite most for B2B data questions?

YouTube and Reddit, ahead of every vendor. Across the 84 AI Overviews Cleanlist collected on August 15, 2026, the most-cited domains were youtube.com (54), reddit.com (35), pipeline.zoominfo.com (23), linkedin.com (16), apollo.io (15), clay.com (15), salesforce.com (14), cognism.com (12) and cleanlist.ai (11). The mean AI Overview in this set cited 9.06 sources, summing to 761 source slots across the 84 overviews. Two user-generated platforms out-citing every vendor marketing site is the most reusable finding in the dataset for anyone planning where to publish.

Can Google cite a page without naming the company that wrote it?

Yes, and Cleanlist has measured the gap in both directions. In the Ghost Citation Report of August 10, 2026, Google cited cleanlist.ai in 30 of 58 answers and wrote the word Cleanlist in 1. In this August 15, 2026 dataset the shape inverts: across 25 Google AI Mode questions, cleanlist.ai was cited in 6 answers and Cleanlist was named in 7. The difference is the query set, since 27 of the 37 August 10 keywords came from Cleanlist's own AI Overview citation gap and were therefore anchored to pages Cleanlist already ranks on, while these 25 were buying questions chosen before any result was seen.

Is this study a ranking of the best B2B data tools?

No. It is a measurement of what Google's answer engines said on one day, and it should not be read as a shortlist. A tool near the top of these tables is a tool Google names frequently, which is a function of how much has been written about it across the web rather than of product quality. Several tools counted here, including Hunter, are Cleanlist data partners whose data Cleanlist buys, and they are counted identically to everybody else because excluding them would corrupt the measurement. Both raw datasets are published under CC BY 4.0 so any recount can be done independently.