The Alternatives Economy: 92 Live SERPs Where Apollo Is Named in Every AI Overview

Last updated: September 1, 2026. This page was revised the same day after an internal fact-check; the corrections are itemised near the end.

We pulled live Google SERPs on September 1, 2026, four query forms across 23 GTM and B2B data vendors, and recomputed every number below from the raw results. All 92 carry an AI Overview. All 92 name Apollo, while apollo.io's own site ranks in the top 10 on 20.7% of them. The vendor you searched for ranks in its own top 10 on 62.0% of its queries and is missing from the first 19 results entirely on 19.6%. Cleanlist competes in this market: we run 16 vendor alternatives pages, so treat everything here as measured by an interested party, and check the CSV.

Where does Cleanlist lose in this study?

Badly, and in two places. First, naming. Google's AI Overviews name Apollo on 92 of 92 alternatives queries and Cleanlist on 2, which is 2.2%. Second, economics. Our 16 live alternatives pages took 614 clicks from 349,145 impressions over the 16 months to August 29, 2026, a 0.176% click-through rate, which works out to about 38 organic clicks a month across all 16 pages combined. The internal figure we had been repeating was roughly 100 a month; the export says 38. We hold position 1 on three of the four RocketReach queries in this study and we are cited in none of those four AI Overviews. Competitors' pages beat ours on reach by a wide margin.

Do Google AI Overviews appear on "[vendor] alternatives" queries?

On every one we pulled. 92 of 92, which is 100%, on September 1, 2026 in the United States on desktop. The four query forms are identical here: 23 of 23 for alternatives, competitors, best [vendor] alternative and alternatives free. Each answer named a mean of 5.7 vendor brands, with a floor of 2 and a ceiling of 8. For a GTM leader the practical reading is that the alternatives SERP is now an answer with a list of links underneath it. Google's own documentation notes that AI features are generated and can vary between users and over time, so treat 100% as "effectively always" rather than as a constant.

Which brand do AI Overviews name most often on alternatives queries?

Apollo, on 92 of 92. It holds even on queries where Apollo is a strange recommendation, including "skrapp alternatives", "lemlist alternatives" and "salesloft competitors". The rest of the leaderboard, counted once per answer, with the full table in section 2 of the dataset:

That table records what Google's answers said, not what we think of any product in it. Every brand is counted by the same rule, the full 43-brand vocabulary ships with the dataset, and no comparison between two tools is drawn anywhere on this page.

Our companion AI Overview citation study of 206 GTM buyer queries puts Apollo at 63.5%, which is 47 of the 74 answers in that corpus that returned a reference list. On alternatives queries specifically it is 100%. The two collections differ in one respect that explains most of the gap: this study set load_async_ai_overview and that one did not, so this study reads every answer while that one reads only the 38% that loaded synchronously. Even like for like, the alternatives SERP is where a single incumbent name gets reinforced hardest.

Does the vendor you searched for rank for its own "alternatives" query?

Less often than most vendors assume. Across all 92 queries the searched vendor's own domain appears in the top 10 on 57 queries (62.0%), in the top 3 on 23 (25.0%), and at position 1 on 7 (7.6%). It is absent from the entire first 19 results on 18 queries (19.6%). When it does rank, its median position is 5. So on more than a third of these queries, a buyer searching for a way off your product sees ten results and none of them is you. The four query forms differ more than we expected: the vendor is absent on 3 of 23 competitors queries, 4 of 23 alternatives, 4 of 23 alternatives free, and 7 of 23 best [vendor] alternative, which is the hardest form to hold.

Which vendors are absent from their own alternatives SERP entirely?

Two of the 23: Outreach and Salesloft returned no result from their own domain on any of their four queries, inside the first 19 positions. At the other end, 11 of 23 held a top-10 position on all four: Amplemarket, Apollo.io, Cognism, FullEnrich, Kaspr, Ocean.io, Persana, Seamless.AI, Skrapp, Snov.io and UpLead. Read this as a measurement of search results on one day and nothing more. It says nothing about product quality, revenue or customer satisfaction. What it does say is that defending your own brand-plus-alternatives query is a choice, and 12 of 23 vendors have not fully made it.

Who holds position 1 on a "[vendor] alternatives" search?

An independent publisher, most often. Across the 92 position-1 results: independent publishers 42 (45.7%), rival vendors 23 (25.0%), review sites 15 (16.3%), the searched vendor itself 7 (7.6%) and community or UGC pages 5 (5.4%). The single most frequent position-1 domain is saleshandy.com with 11, then gartner.com with 9, rb2b.com with 7, and sparkle.io and g2.com with 6 each. None of those is the vendor being searched for. The top of this SERP is owned by whoever decided to write the page, and in 92% of cases that is somebody other than the company whose brand is in the query.

How often does a rival vendor take position 1 on a competitor's query?

On 23 of 92 queries, which is 25.0%. The most successful land-grabber in the sample is saleshandy.com with 11 position-1 results, spread across queries targeting six different vendors. Behind it, three domains hold 3 each: cleanlist.ai, all of them RocketReach queries; getprospect.com, all of them Skrapp queries; and pipeline.zoominfo.com, spread across two vendors. The pattern worth noticing is the second one. A challenger picks a single competitor and takes every form of that competitor's query rather than spreading thin. Whether it is worth doing is a separate question, answered further down.

Which domains win the most "[vendor] alternatives" rankings?

Counting top-10 appearances across the 92 queries: reddit.com 79 appearances on 75 distinct queries (81.5% of the set), pipeline.zoominfo.com 44 on 44, sparkle.io 37 on 33, salesforge.ai 35 on 35, saleshandy.com 35 on 35, g2.com 30 on 29, cognism.com 28 on 28, snov.io 27 on 18, listkit.io 20 on 20 and warmly.ai 20 on 20. The market is less concentrated than the position-1 table suggests: 138 distinct domains hold at least one top-10 slot, the top 10 domains hold 38.6% of all top-10 slots, the top 25 hold 62.6%, and 37 domains (26.8% of those that appear at all) appear exactly once. There is still room at the bottom of page one, and almost none at the top.

Which domains do AI Overviews cite on alternatives queries?

The published dataset carries one row per organic result, so it can only mark a citation on a domain that also ranked somewhere in the first 20 results of that query. On that basis there are 372 unique query-and-domain citations across 87 domains: sparkle.io on 24 queries (26.1%), g2.com 24, pipeline.zoominfo.com 24 and apollo.io 24, then reddit.com 16 (17.4%), derrick-app.com 15, syncgtm.com 13, warmly.ai 13, enrich.so 12 and youtube.com 11. cleanlist.ai is cited on 6 (6.5%).

By page type, that citation pool splits: independent publishers 187 (50.3%), rival vendors 113 (30.4%), review sites 30 (8.1%), community pages 28 (7.5%) and the searched vendor's own site 14 (3.8%). The vendor's own defence is the least-cited source type in the whole set. Google reads the third-party write-ups about you.

One limitation you should hold against this section. A domain Google cited but never ranked has no row in this file to be marked on, and the per-answer reference lists are not in the published dataset, so the leaderboard above is a floor rather than a full count. The first version of this page published a fuller citation leaderboard built directly from the reference lists. We have withdrawn those figures rather than publish numbers nobody can check against the file we shipped. Everything below is recomputable from the CSV.

Do the domains cited by the AI Overview also rank on the page?

Mostly not, and both directions of that are visible in the file. 127 of the 372 citations (34.1%) go to a domain that ranked only past position 10 on that query, so it was never on page one for the buyer who asked. Pushing the other way: of the 880 distinct domain-and-query pairs holding a top-10 slot across this set, 635 (72.2%) were never cited by the AI Overview sitting above them. Both of those are floors, because the domains Google cited without ranking at all are invisible to this file entirely. This confirms on a second, differently-built dataset what our ghost citation work found earlier. A domain can win the ranking competition and lose the citation competition on the same query, and 635 times in this dataset it did.

Do the brands named in the answer have a page ranking for that query?

Usually not. Of the 526 brand-namings that map to a known vendor domain, 381 (72.4%) name a vendor whose own website does not appear in that query's top 10. Only 145 (27.6%) name a vendor that also ranks. The registry that makes that computable is section 2 of the dataset: 43 brands, one primary domain each, with the naming count and the ranking count per brand so you can rebuild the figure or disagree with a mapping. Apollo is the extreme case: named in 92 of 92 answers, cited as a source on 24, and ranking in the top 10 on 19. Google is willing to recommend a brand it has not read on that query and has not ranked on that query. The mechanism is presumably that the third-party pages it did read all say the same name. For a challenger brand, that is either bad news or the whole strategy, depending on whether the pages that get read are pages you can influence.

How often does Reddit appear on a "[vendor] alternatives" SERP?

On 75 of 92 queries in the top 10 (81.5%), and on 28 in the top 3 (30.4%). It is the single most frequent domain in the study by a distance. We cannot score Reddit's page shape, because all 11 Reddit URLs we fetched returned a client-rendered shell of under 300 words to a plain HTTP request, so they are excluded from every page-shape figure below. That exclusion is itself worth stating: the most common result type on this SERP is the one our page-shape method cannot read, and any claim we make about "what a ranking page looks like" is a claim about the non-Reddit results only.

How long is a page that ranks for "[vendor] alternatives"?

Long. Across the 67 analysable pages the median is 3,851 words and the mean is 4,223, with a floor of 536 and a ceiling of 13,706. Only 5 pages (7.5%) come in under 1,500 words, and 17 (25.4%) run past 5,000. Splitting by position, the median for position 1 is 3,851 words, for position 2 it is 4,632, and for position 3 it drops to 2,998.5 (an even sample of 18, so the median is a midpoint). Length is not the causal lever here and we are not claiming it is; what the distribution shows is that the entry ticket to this SERP is a substantial page rather than a 900-word landing page with a feature table. The word count is computed on visible text after stripping script, style, SVG and noscript blocks.

How many alternatives does a ranking page actually list?

A median of 9, a mean of 8.4, a maximum of 17. The counting rule: we take every h2 and h3 on the page and count how many distinct vendor names from a fixed 123-name dictionary appear in that heading text. 27 of 67 pages (40.3%) list 10 or more. Ten pages (14.9%) list two or fewer, and those are almost entirely the vendor's own defensive page, where the "alternatives" page is really a product page. This rule undercounts pages that list vendors in a table without giving each one a heading, and it overcounts any page with vendor names in a navigation heading. Both failure modes are visible in the data, so read 9 as the central tendency of a noisy measure rather than a precise figure.

Do alternatives pages show prices, and do they date them?

Prices appear on 51 of 67 pages (76.1%), where the rule is at least three distinct currency-amount tokens in the visible text. Of those 51, 36 (70.6%) date the price in some way, either with an explicit marker such as "as of" or "last updated" or by placing a year between 2024 and 2026 within 300 characters of a price. Against the whole sample that is 53.7%. So roughly a quarter of ranking pages quote no prices at all, and of the pages that do, three in ten give the reader no way to know when the number was true. Of the rules we applied to these pages, undated pricing is the most commonly failed one, and it is the cheapest to fix.

Do alternatives pages disclose that the publisher is a competitor?

Almost never. 3 of 67 pages (4.5%) contain a bias disclosure, matched against a fixed phrase list ("full disclosure", "disclosure:", "we built", "we are biased", "conflict of interest" and similar). Narrowing to the 35 pages published by a vendor with a commercial stake in the answer, 2 disclose, which is 5.7%. Credit where it is due: one of the two is skrapp.io, which writes "Full disclosure: Skrapp's been around for years for a reason". Cleanlist's two pages in the sample, /alternatives/rocketreach and /alternatives/fullenrich, carry no disclosure at all. We are 2 of the 33 pages in the non-disclosing majority, and we have no basis for exempting ourselves from the criticism.

Do alternatives pages say anything positive about the incumbent?

This one resisted automated measurement, and the honest thing is to publish all three attempts. A loose rule (any positive adjective within 250 characters after the incumbent's first mention) returned 13.4%. A tighter rule with listicle-title language stripped returned 83.6%. Reading the evidence, both were unreliable, because the adjective frequently described a different tool in the same window. The rule we settled on requires a sentence where the incumbent is the grammatical subject ("Cognism is", "Apollo offers", "Clay has"), carrying a positive term and no negation. On that rule, 24 of 67 pages (35.8%) say something clean and positive about the incumbent. Among the 54 pages with any such sentence, 24 are positive and 30 are negative. Rival vendors are more generous than independent publishers: 40.7% versus 25.0%.

Structurally the picture is friendlier. 56 of 67 pages (83.6%) give the incumbent its own h2 or h3 section rather than treating it only as the thing being escaped, rising to 88.9% on rival-vendor pages.

What does a page that ranks for this actually look like?

Composited from the 25 position-1 pages in the sample: about 3,900 words, 9 named alternatives each with its own heading, a comparison table (68.0% have one), and prices (92.0% show them). Position 3 looks materially thinner: a 2,998.5-word median, a comparison table on 33.3%, prices on 50.0%. The strongest single differentiator between position 1 and position 3 in this sample is the comparison table and the presence of prices, not length. Two caveats on that: n is 25 and 18, and this is correlation on one snapshot with no control for domain authority, so treat it as a description of what is up there rather than a recipe for getting there.

Does ranking #1 for "[vendor] alternatives" still bring traffic?

Here is the one first-party data point that started this study, and it is an anecdote rather than a law. In our own Search Console export covering April 30, 2025 to August 29, 2026, the query rocketreach alternatives returned 57 clicks from 373,692 impressions, a 0.0153% click-through rate, at an average position of 3.01. Today's live SERP, pulled for this study, puts cleanlist.ai at position 1 for that query, position 1 for "best rocketreach alternative", position 1 for "rocketreach alternatives free" and position 2 for "rocketreach competitors". On all four, the AI Overview cites domains that also rank on the query, and on none of the four is one of them ours.

Treat the impression figure with suspicion. The page /alternatives/rocketreach reports 162,929 impressions over the same window, less than half the query-level figure, and Search Console aggregates queries and pages separately in ways that do not reconcile. Our companion Search Console study works that inconsistency through in detail. What survives is a question rather than a finding: position 1 on this SERP no longer implies the traffic that position 1 used to imply, and anyone modelling the ROI of an alternatives page on a published CTR curve is probably modelling the wrong thing.

What did Cleanlist's own alternatives pages actually produce?

Sixteen live vendor pages, plus an /alternatives index, so 17 URLs. Over the same 487-day window they took 614 clicks from 349,145 impressions, a 0.176% CTR, which is 38.4 clicks a month across all 16 pages combined. Including the index it is 38.6. Two pages clear 100 clicks for the window, /alternatives/apollo at exactly 200 from 54,710 impressions and /alternatives/rocketreach at 154 from 162,929, and one page (/alternatives/cognism) took zero. Against site totals in that export of 40,763 clicks and 7,725,528 impressions, the 16 live pages are 1.51% of clicks and 4.52% of impressions. In Google's AI-features export for May 18 to August 29, 2026, the same 16 pages are 9.66% of our AI impressions (63,493 of 657,591). That makes the cluster 2.14x more concentrated on the AI surface than on the ordinary search surface, while in absolute click terms it stays close to a rounding error.

Do the four query forms behave differently?

Modestly. AI Overview coverage is 100% on all four. The mean number of brands named per answer runs competitors 6.5, alternatives 5.9, best [vendor] alternative 5.7, [vendor] alternatives free 4.8, so the free-modifier query produces the shortest shortlist. Vendor self-presence in the top 10 runs 16 of 23 for competitors, 15 of 23 for alternatives free, and 13 of 23 for alternatives and best [vendor] alternative. The practical consequence is that a single well-built page can plausibly serve all four forms, which is what the RocketReach and FullEnrich clusters show in our own data, where one URL holds three or four of them at once.

What did this page get wrong in its first version?

Five corrections, all made on September 1, 2026, before this page was linked anywhere.

The target set shrank. Vendors with a commercial relationship to Cleanlist were removed from the query set, which took the study from 120 queries and 2,244 rows to 92 queries and 1,727 rows across 23 vendors. Every SERP-level figure on this page has been recomputed on the published file rather than adjusted. The findings that moved: the vendor's own top-10 rate 66.7% to 62.0%, its position-1 rate 10.0% to 7.6%, rival vendors at position 1 28.3% to 25.0%, the top-10 domain concentration 37.2% to 38.6%, Reddit's top-10 rate 75.0% to 81.5%, and Cleanlist's own naming rate 2.5% to 2.2%. The finding that did not move at all: every single answer in the set carries an AI Overview and every single one names Apollo.

The companion study was misdescribed and its wrong figure repeated. The first version called it "an AI Overview citation study of 100 general B2B buyer queries" and put Apollo at 24.6%. It is 206 GTM buyer queries, not 100 general ones, and its 24.6% divided by every detected AI Overview when only the answers that returned a reference list could produce a brand mention at all. On the base that can, which is 74 answers in that corpus, Apollo is 63.5% and Cleanlist is 2.7%. The comparison against this study's 100% is still a real finding and a much smaller one than the page first claimed.

The citation leaderboard was not recomputable from the published file. The first version reported 754 citations from 148 domains, a page-type split, and "409 of 754, 54.2%, come from a domain that does not appear in that query's top 10", all built from AI Overview reference lists that the published CSV does not carry. Those figures are withdrawn. The citation section now reports only what the file supports: 372 query-and-domain citations across 87 domains, a page-type split on the same base, 34.1% of citations going to domains that ranked only past position 10, and 635 of 880 top-10 pairs never cited. Every one of those is a floor.

The top-10 domain list was wrong and one sentence inverted its own fact. The printed top-10 list skipped its true eighth entry. It has been rebuilt from the file. And the disclosure section read "we are 33 of the 35 non-disclosing majority" where the correct statement, given later on the same page, is that we are 2 of the 33.

Two pages rounded the same CTR differently. rocketreach alternatives was published here as a 0.02% click-through rate and in the companion Search Console study as 0.0153%, from the same 57 clicks and 373,692 impressions. This page now uses 0.0153%.

What are the limits of this study?

Six, stated plainly. One snapshot, September 1, 2026, so nothing here establishes a trend. One country and one device, United States desktop; our own Search Console shows desktop at 0.300% CTR against mobile at 2.380%, so a mobile pull would likely tell a different story. AI Overviews are generated and vary, between users and over time, and Google says so; a second pull an hour later would not match this one exactly. This is a parse, not a human read: brand naming, page typing, price dating and incumbent sentiment are all regex rules, each stated above, and each has failure modes we have named. The page-shape sample is 67 pages, excludes all 11 Reddit results, the five domains that returned HTTP 403 to a direct fetch, among them g2.com, gartner.com, softwareadvice.com and crozdesk.com, enrich.so on a 429, and snov.io on a connection error, which biases it toward publishers that permit crawling, and it was collected before the target-set exclusion so it is wider than the published CSV. We are an interested party: Cleanlist runs 16 alternatives pages, ranks in this dataset, and chose the vendor list. Vendors that have a commercial relationship with Cleanlist were excluded from the target set precisely so that none of them is ranked or compared against us here. Brands that Google's own answers happened to name are recorded in the naming table by the same rule as every other brand, with no product comparison drawn anywhere on this page.

Where can I download the data?

The full result-level dataset is at /data/alternatives-serp-economy-2026-09.csv, released under CC BY 4.0. Attribute to Cleanlist with a link to this page.

Section 1 is 1,727 rows, one per organic result. Column dictionary:

Section 2 is 43 rows, the complete brand-to-domain registry behind the naming rule: brand, primary_domain, ai_overviews_naming_it, share_of_92, queries_with_own_domain_in_top10, namings_where_own_domain_also_ranks and namings_where_own_domain_does_not_rank. A brand outside that list scores zero by construction, which is the single biggest source of measurement error in the naming figures, so the list is published in full rather than described.

One caveat on cited_in_ai_overview, because it will not reconcile otherwise. The file has one row per organic result, so a domain that Google cited but did not rank has no row to be marked on. Every citation figure on this page is therefore a floor, and is described that way in the citation section.

To reproduce: POST each query to the DataForSEO Google Organic Live Advanced endpoint with location_code 2840, language_code "en", device desktop, depth 20 and load_async_ai_overview true, then apply the typing and naming rules above. The AI Overview text will differ from ours; the structural findings should not.

What is Cleanlist changing on its own alternatives pages?

Two things, from our own numbers rather than anyone else's. We are dating every price on those pages, because 30% of priced pages in this study leave the reader guessing. And we are re-weighting the cluster toward the AI surface, where it earns 9.66% of our AI impressions against 4.52% of our ordinary ones, and away from a click model that returns 38 visits a month.

For what it is worth on the product side: Cleanlist starts every new workspace on Scale for 14 days, 250 credits, 3 seats, no card, and Free afterwards is $0 for 30 credits a month. Paid plans are Starter $79, Pro $229 and Scale $599, 25% off annually. Search costs 0 credits, a verified email is 1, a phone is 10. Our 500-lead enrichment benchmark returned 98% verified email and 85% direct dial through the 25+ provider waterfall. The 16 pages this study measures are our alternatives hub, and the buyers who land on them are the outbound and RevOps teams this research is written for.