Breezy HR shows up nearly everywhere, except when the question shrinks
Breezy HR appeared in five of the six answers that ChatGPT, Gemini, and Perplexity returned when asked about applicant tracking systems. Had we stopped at five general or affordability-focused questions, Breezy HR would have looked like the assistants’ default pick. The single question that excluded it, “best applicant tracking systems for small teams,” yielded an entirely different set of names, many of which never showed up elsewhere.
All six answers drew from reddit.com and technologyadvice.com. Those two domains fed every assistant response, regardless of phrasing. Quora and zapier.com appeared in four of the six, and peoplemanagingpeople.com, theundercoverrecruiter.com, and vendordirectory.shrm.org each appeared in three. The recommendations, in other words, were being built from a very small pool of sites. BambooHR and Greenhouse each landed in four answers, but with different coverage gaps. BambooHR was named in “best applicant tracking systems,” “best for small teams,” “what applicant tracking systems should I use,” and “most affordable.” Greenhouse appeared in “best,” “what should I use,” “most affordable,” and “applicant tracking systems recommendations.” Neither appeared in the free question.
When the question became “best applicant tracking systems for small teams,” the assistants named BambooHR, 100Hires, JazzHR, AvaHR, Zoho, VIVAHR, Talexio, and Enginehire. 100Hires, VIVAHR, Talexio, and Enginehire were not mentioned in any other answer. Breezy HR, which showed up in the free and affordability questions, vanished. That suggests the assistants treat team size as a hard cut, not a filter on a broader list, and that they surface a completely distinct set of vendors for small-team buyers.
Across six questions, the three assistants collectively named 22 unique products. Eight products appeared in three or more answers: Breezy HR (5), BambooHR (4), Greenhouse (4), Ashby (3), JazzHR (3), Manatal (3), Pinpoint (3), and Zoho Recruit (3). The rest were thin. Fourteen products appeared just once, including iCIMS, Workable, and Lever, which are well-known in the ATS market but barely surfaced here. For a seller, the concentration means that if your product isn’t in the top handful of names these assistants repeat, you’re effectively invisible across a range of purchase-intent queries.
Why Breezy HR missed the small-team answer despite appearing in “best free” and “most affordable” is not obvious from the source list. The small-team question drew on uschamber.com, g2.com, and enginehire.io alongside the usual reddit and technologyadvice.com, but the inclusion logic remains opaque. This measurement cannot tell us whether the assistants interpreted “small teams” as a distinct buyer profile or simply pulled from a different set of articles that omitted Breezy HR. That gap is something any vendor tracking these queries would need to dig into, because it suggests the assistants are not simply reranking the same shortlist for every variant.
Three products hit four mentions, none hit six
No applicant tracking system appeared in every single answer. Breezy HR came closest, showing up in five of the six responses across ChatGPT, Gemini and Perplexity. That near-sweep makes it the de facto default recommendation when these assistants field a general or affordability-oriented ATS question. But the one question that skipped it, “best applicant tracking systems for small teams,” pulled in a completely different roster, so even Breezy HR’s dominance has a clear boundary.
Two names tied at four mentions: BambooHR and Greenhouse. Their pattern is tight. They both appeared in the broad “best” question, the consultative “what should I use,” and the two affordability prompts, then dropped out only for the small-team and free-product queries. That pairing suggests the assistants treat them as the safe, full-feature mid-market picks, less relevant when cost or team size tightens the frame.
A cluster of five tools landed at three mentions: Ashby, JazzHR, Manatal, Pinpoint and Zoho Recruit. Their distribution differs. Ashby showed up for “best,” “what should I use,” and “recommendations,” meaning it never made it into an answer that specifically asked for free or affordable systems. JazzHR belonged to “small teams,” “what should I use,” and “most affordable,” a mix of price sensitivity and scale. Manatal and Pinpoint both claimed spots in “best,” “most affordable,” and “recommendations,” while Zoho Recruit appeared in “best free,” “most affordable,” and “recommendations.” Zoho Recruit carries an extra distinction: the standalone product name “Zoho” without “Recruit” was named only once, in the small-team answer, as a separate entity. The assistants may be conflating the parent brand with the ATS module, but the count treats them as different entries, and that single mention of plain “Zoho” is fragile.
Two products reached two mentions: AvaHR and Lever, both enough to escape the noise of the long tail. AvaHR was listed for small teams and free systems, a pairing that makes sense for a tool pitched at startups. Lever appeared under “best” and “what should I use,” aligning with its reputation as a recruiting-suite product. Teamtailor also got two mentions, both in the broad best and the recommendation prompt, so it lands closer to the Ashby tier in positioning.
Eleven products were named once and only once. That list reads like a directory of niche, newer, or position-specific tools: 100Hires, Applisights, Enginehire, Gem, iCIMS, JuggleHire, Loxo, MightyRecruiter, Talexio, Tracker, VIVAHR. Workable and the bare “Zoho” also belong here. A single mention across six questions is not a recommendation pattern, it is a fragment. For the vendor, it means that in one specific phrasing, one assistant pulled their name; for a buyer, it means they would likely never see these names if they asked the other five questions.
What stands out most in the single-mention column is iCIMS. A well-known enterprise ATS with a large installed base, it received exactly one mention, in the “what should I use” answer. That gap is more striking than the absence of midsize players because iCIMS is often synonymous with high-volume recruiting. It appears that the assistants’ sourcing patterns,heavily drawn from reddit.com, technologyadvice.com, and a cluster of review and listicle sites,favored tools with strong presence in those forums over a tool more commonly discussed in procurement reports.
Those same sourcing constraints kept out entire classes of applicant tracking systems. No answer named any large-suite HCM platform that bundles ATS functionality, even though many buyers start their search there. The absence of those names from our six questions does not signal that the assistants consider them inferior; it signals that the assistants’ retrieval chains rarely surfaced them for these concrete, comparison-style prompts. Buyers who need to choose between an integrated suite and a standalone ATS got no help from these answers. That gap is worth knowing before treating an assistant’s list as a market map.
The full list, counted
Every product serper+or:chatgpt,gemini,perplexity named across 6 answers, with how many of those answers named it. The share column is out of the answers we read, not out of the category.
| Product | Named in | Share | Read as |
|---|---|---|---|
| Breezy HR | 5 of 6 | 83% | 5 answers |
| BambooHR | 4 of 6 | 67% | 4 answers |
| Greenhouse | 4 of 6 | 67% | 4 answers |
| Ashby | 3 of 6 | 50% | 3 answers |
| JazzHR | 3 of 6 | 50% | 3 answers |
| Manatal | 3 of 6 | 50% | 3 answers |
| Pinpoint | 3 of 6 | 50% | 3 answers |
| Zoho Recruit | 3 of 6 | 50% | 3 answers |
| AvaHR | 2 of 6 | 33% | 2 answers |
| Lever | 2 of 6 | 33% | 2 answers |
Two domains decided every recommendation, and neither was a vendor
Two domains fed all six answers across ChatGPT, Gemini and Perplexity: reddit.com and technologyadvice.com. A forum and a review-comparison site. That pairing sat at the centre of every response, no matter the phrasing. Quora and zapier.com were close behind, feeding four of the six. No other domain broke three appearances. The group at three included peoplemanagingpeople.com, theundercoverrecruiter.com and vendordirectory.shrm.org, a mix of an HR publication, a recruitment blog and the SHRM vendor directory.
Vendor domains, when they did show up, appeared once and only once. lever.co, enginehire.io, gem.com, jugglehire.com, loxo.co. Each surfaced inside a single answer linked to a specific question. Enginehire.io, for example, fed the “best applicant tracking systems for small teams” response, and Enginehire was named in that answer. Loxo.co fed the “best free” question, which then listed Loxo. The pattern is tight: vendor pages only entered when the question’s phrasing happened to trigger a source that matched a named product, and they never repeated across questions. Contrast that with reddit, which was present even in the answer that pulled from six other domains alongside it.
The questions themselves shuffled the supporting cast. The broad “best applicant tracking systems” query drew from technologyadvice.com, reddit.com, vendordirectory.shrm.org, peoplemanagingpeople.com, lever.co, youtube.com and myshortlister.com. A standard spread: one vendor, one directory, several content sites, one video platform. “Best free applicant tracking systems” dropped youtube entirely, brought in quora.com and giighire.com, and pulled from loxo.co and peoplemanagingpeople.com. That answer still had reddit and technologyadvice in it. The small-team question was the odd one out, sourcing uschamber.com, g2.com, enginehire.io and zapier.com alongside the two anchors. g2.com appeared in no other answer, and uschamber.com only showed up in one other, the “most affordable” question. The source divergence matches the product divergence: Breezy HR, present in every other answer, vanished from the small-team list.
What this mix says about how a recommendation gets made is clear enough to be uncomfortable. The assistants do not build answers by crawling vendor feature pages or pricing sheets. They build them by leaning on aggregated discussion and comparison content, the kind of material that gets linked in round-ups and forum threads. Reddit and technologyadvice.com are everywhere because they are that material. When a vendor domain breaks through, it is a single-answer event, tied to a narrow question and not repeated.
Why technologyadvice.com achieved a perfect six while g2.com, arguably a larger review platform, showed up once, is not something this measurement can explain. The same gap exists for Quora at four versus facebook.com at one. The data only shows that an assistant’s recommendation set is, in large part, a function of which comparison and discussion pages happened to anchor its retrieval for a given phrasing. Change the phrasing, and you change which secondary domains get pulled in, which in turn reshapes the product list. The two anchors stay fixed, but around them, the category is decided by a shifting handful of review sites, forums and, rarely, a lone vendor page.
The source trail explains the product list, and leaves gaps
Two domains, reddit.com and technologyadvice.com, fed all six answers. That is the simplest explanation for why Breezy HR landed in five of them. A product that shows up on both an enormous forum and a site that specialises in software comparisons does not need to win on features. It just needs to be present in the text that the assistant ingests. Quora and zapier.com appeared in four of six answers, so the typical response leaned on a tight bundle of four sites: Reddit, TechnologyAdvice, Quora and Zapier. Any applicant tracking system mentioned across those four, especially if it appeared repeatedly in the threads or listicles that get surfaced, had a strong chance of being recommended.
BambooHR and Greenhouse both hit four mentions, and their pattern maps cleanly onto that source bundle. They were named in the broad “best” question, the more personal “what should I use” question, the “most affordable” question, and the straight “recommendations” question. Each of those answers drew on some combination of Reddit, TechnologyAdvice, Quora and Zapier. In the affiliate-heavy, list-driven world of “best ATS” content, BambooHR and Greenhouse are almost furniture. When an assistant’s retrieval step pulls from those domains, it pulls the products those domains feature. The same logic holds for Manatal, Pinpoint and JazzHR, each named three times. Their appearances cluster in the exact same questions and the same source set.
Ashby, at three mentions, is an interesting case. It appeared in the general “best” question, the “what should I use” question, and the “recommendations” question, all of which were fed by Reddit and TechnologyAdvice plus extras like vendordirectory.shrm.org and theundercoverrecruiter.com. Ashby is a newer, startup-favoured product, and Reddit discussions about recruiting tools surface it frequently. The assistant does not need to know what Ashby does. It only needs to retrieve a thread where someone named it, and that thread becomes the recommendation. The fact that Ashby did not appear in the “affordable” or “free” questions suggests those specific source sets (which included uschamber.com, giighire.com, loxo.co) simply did not mention it, or mentioned it in a different context that the assistant did not surface as relevant.
Then the question “best applicant tracking systems for small teams” upends everything. The sources shift: uschamber.com, reddit.com, g2.com, peoplemanagingpeople.com, technologyadvice.com, zapier.com, and enginehire.io. That mix introduced a slate of products that appeared nowhere else: 100Hires, AvaHR, Talexio, VIVAHR, and Enginehire itself. Enginehire’s own domain showed up only in that question, and Enginehire the product only in that question. This is a direct line from source to recommendation. The assistant saw a vendor site among the retrieved pages and treated its name as a credible answer. The same pattern holds for Gem, which appeared once, in the “what should I use” question, alongside gem.com in the source list. Loxo appeared once, in the “best free” question, with loxo.co in the sources. When a vendor domain gets pulled in, that vendor becomes the answer. We cannot tell whether the assistant would have named these products otherwise, because the data gives us no control. The retrieval step already decided the candidate set.
What we cannot see is the content of the pages that were actually used. Technologyadvice.com might have a list of the top ten ATSs, but we do not know whether the assistant grabbed a 2024 article or a 2023 one, whether it read the full page or a snippet, or whether it weighted certain products based on how they were described. Reddit threads are episodic: a single mention of a product in a heavily upvoted comment might be enough. Quora answers vary wildly in quality. The assistant’s recommendation is the output of a black-box retrieval-plus-generation pipeline, and all we observe is the final list of names and the domains that fed the pipeline. That means there could be a highly suitable ATS that never appears on these four or five domains. It would be invisible to the assistant. The measurement tells us what got recommended given this particular internet, not what could or should have been recommended.
Your position is actually a handful of mentions on two or three sites
If you sell an applicant tracking system and your product appeared in this measurement, you are not defending a reputation. You are defending a scattering of mentions across reddit.com, technologyadvice.com, and perhaps quora.com or zapier.com. That is the entire foundation. Breezy HR, the most-recommended product with five of six answers, owes that lead to being present in the threads and listicles those four domains surfaced for almost every question. The single question that skipped it, “best applicant tracking systems for small teams,” pulled its sources from uschamber.com, g2.com and enginehire.io instead. Breezy HR had no foothold there, so it vanished from the answer entirely. One different phrasing, one different bundle of source pages, and the lead evaporated.
That is what fragility looks like. BambooHR and Greenhouse, each named in four answers, ride on the same tight source cluster. Both appeared for the broad “best” and “what should I use” questions, but neither showed up in the free-ATS answer. A single editorial update to a technologyadvice.com roundup, or a shift in which Reddit thread the assistant retrieves, and they become three-mention names overnight. Losing one mention does not drop them out of the dataset. It drops them into a crowd of six other products, all sitting at three mentions, where the signal is no longer “this is a top pick” but “this is one of many.”
That three-mention tier, Ashby, JazzHR, Manatal, Pinpoint and Zoho Recruit, is the most precarious of all. These names are present, but barely. Their appearance pattern is inconsistent. Ashby shows up in the broad “best” and the consultative “what should I use,” then disappears in affordability and small-team queries. Pinpoint is tied to affordability and recommendations, absent elsewhere. The difference between a three-mention product and a singleton like Workable, which showed up only in the recommendations question, is often a single Zapier article or a single Reddit comment. If that piece of content gets replaced, you fall off the list the assistant builds. The assistant does not reason that Ashby is a strong generalist. It compiles what the query retrieves.
Products with two mentions, AvaHR and Teamtailor, face the same dynamic with even less margin. Lever, named twice, had its own domain, lever.co, feed one answer. That is a tiny thread. If lever.co were no longer in the crawl, or if the assistant’s retrieval mechanism changes, that mention would likely disappear. Vendors with a single mention are effectively invisible to the assistant for all but the specific query that happened to pull them in. They are not a recommendation. They are artefact.
The cost of losing your position is not a multi-year branding effort. It is a source page being re-ordered, a new Reddit thread outranking an old one, or a competitor managing to get named in the same technologyadvice.com list. The assistants do not have memory. They reconstruct an answer from whatever pages score highest for that query at that moment. A product that today appears in five of six answers could appear in two next month if the core sources refresh their content and that product’s mention is cut. Breezy HR’s absence from the small-team question already shows the mechanism. Defending a position means monitoring those four key domains obsessively and ensuring your product’s name appears on them, in the right context, for each query variant you care about. Otherwise, you are one domain refresh away from being replaced by a competitor who understood the same simple rule.
If you were not named, your absence is in a handful of pages, not in your features
ChatGPT, Gemini, and Perplexity did not run a test drive. They did not assess your workflow speed, your Slack integration, or the depth of your reporting. Across six questions about applicant tracking systems, four domains supplied the raw material for nearly every product mention: reddit.com, technologyadvice.com, quora.com, and zapier.com. If your product did not appear in a single answer, the reason is almost certainly that it does not appear, or does not appear prominently enough, in the threads, listicles, and comparison tables hosted on those four sites.
That is a very narrow surface to defend or attack. The assistants ingested a tight bundle of pages, often the same URLs recycled across questions. When reddit.com and technologyadvice.com alone fed all six answers, a product missing from both is invisible to the system. No amount of customer satisfaction, no analyst report, and no paid campaign elsewhere will bridge that gap, because the assistants are not pulling from a broad index. They are pulling, via search retrieval, from whatever pages rank for each query, and for this measurement those pages came overwhelmingly from the same small set of domains.
Getting into the answers, then, means getting into those pages. On Reddit, that involves earning mentions in the threads that surface for phrases like “best applicant tracking systems” or “ATS recommendations.” It is not about creating a subreddit; it is about being the product someone types out in a comment when a user asks “what should I use.” On TechnologyAdvice, it means being included in the comparison articles and roundups that the site publishes, and often that requires meeting their inclusion criteria, which may hinge on market presence, feature set, or reviewer access. Quora works similarly to Reddit: organic, user-generated answers that name you. Zapier’s ATS articles, if they continue to rank, are another placement point, though Zapier appeared in only four of six answers.
The small-team question is a separate door. It pulled from uschamber.com, g2.com, and enginehire.io. A product that sells to small teams but showed up nowhere in this measurement is absent from those sources too. G2.com, for instance, appeared in only one answer, but it fed that specific small-team response. If small teams are your core, building a presence on G2’s ATS category pages and in Chamber of Commerce roundups is not an abstract branding exercise. It is a direct line to being named by ChatGPT, Gemini, and Perplexity when someone asks that exact question.
There is no alternative route. The measurement shows zero cases where an assistant pulled a product name from a vendor’s own website and placed it into a recommendation. Lever.co, Gem.com, Enginehire.io all appeared once each, and only as sources, not as products sourced from their own domain. You cannot write a blog post on your site and expect the assistant to surface it unless that blog post first ranks in the retrieval step and then gets treated as an authoritative source, which did not happen here. The whole mechanism runs on third-party pages, and the third-party pages that matter are surprisingly few: a forum, a tech review site, a Q&A platform, and an automation blog, plus a handful of others for niche queries.
If you sell in this category and were not named, the work is not better features, more advertising, or a new pricing page. It is a precise, labor-intensive effort to get your name into the specific articles and threads that these assistants read. That means identifying which Reddit threads are retrieved for each ATS question, understanding why certain products get mentioned in them, and finding a legitimate way to become part of those conversations. On review sites like TechnologyAdvice or G2, it means getting listed and then maintaining a profile strong enough to be included in editorial picks. None of that is fast, and none of it comes with a guarantee, but it is the only path the data offers.
Pick six questions, open three tabs, and start copying
Pick six questions that vary the intent without drifting into a neighbouring category. The set that drove this measurement was “best applicant tracking systems,” “best applicant tracking systems for small teams,” “what applicant tracking systems should I use,” “best free applicant tracking systems,” “most affordable applicant tracking systems,” and “applicant tracking systems recommendations.” Replace the category with your own. If you sell email verification tools, the first becomes “best email verification tools” and the second “best email verification tools for small teams.” Do not overthink the exact phrasing. The point is to cover a general recommendation, a buyer-fit question, a free tier question, a price-sensitive question, and one that sounds like someone asking colleagues.
Open ChatGPT, Gemini and Perplexity in separate browser tabs. For each question, paste it into each assistant and copy the full answer into a plain text file or a spreadsheet. Three assistants, six questions, eighteen answers. The model running inside each assistant will drift over time, so do not spread this across days. An hour’s typing and copying is enough. Use the default model each assistant hands you. Do not tweak settings.
The real work starts after you have the answers. You need to extract every product or vendor name that the assistant actually recommended, not every product mentioned in passing. Then extract every source domain. With the assistants that show clickable citations, you can grab the domain from the link. With others you will need to read the answer and recognise where a domain is being cited, like “according to a thread on reddit.com.” That part is manual and slow. You will scroll, highlight, type. Often the same product appears twice in one answer under different spellings. You will correct that as you go.
Count occurrences per product across the six questions, per assistant. Count source domain appearances the same way. A spreadsheet with columns for question, assistant, products, and sources lets you pivot later, but you still have to fill it cell by cell. Eighteen answers might yield forty distinct product names and thirty different domains. Double-counting a product because an assistant called it “Breezy HR” in one sentence and “Breezy” in the next is easy to do. Running a quick sort at the end will show you that you made the mistake about an hour after you could have fixed it without recounts. That is the tedium: it is not clever, it is not automatable with the tools you have open right now, and it requires just enough attention that you cannot do it while half-listening to a call.
Once you have the counts, you have what we had. You can see which product shows up in five of six answers, which domain fed all six, and where the gaps sit. That is the analysis from the earlier sections, and it holds for any category as long as you ran the same questions.