Services
Company
Industries
Knowledge
Projects
UK
.png){kind=small}
An impression-weighted analysis of AI Mode and AI Overview queries entering five fintech sites, plus an intent-inference layer to separate buyer queries from learner queries.
Data source. Google Search Console keyword data segmented to AI Mode and AI Overview queries, drawn from five fintech project accounts on the Atomic AGI platform. All five sites are anonymised throughout this report and referred to as Sites A through E. Calendar-month window: June 1 – June 30, 2026. 404 queries in total, 251,835 total impressions.
What gets analysed. Non-branded queries only. Each project's queries are filtered against that site's own brand terms before pooling. After per-site branded filtering: 385 non-branded queries, 250,880 impressions.
Two segments. The five sites split into two clear segments by primary audience:
Impression weighting. Each category's share is calculated by total impressions, not by query count. This is the same weighting used in Foundation Study #1 and remains the closest available proxy to actual demand frequency in AI search.
Statistical approach. Where percentages are reported at the segment level, we compute 95% Wilson score confidence intervals on impression-weighted proportions with the same method as Foundation #1. Wilson intervals are used because they behave correctly at small and extreme values, which matters here because the smallest categories in the B2B pool sit below 2%.
Audience limitation. This dataset captures queries entering AI Overviews and AI Mode via the sites in the pool. It does not identify who typed those queries. The people typing queries about fintech topics into AI search could be buyers evaluating tools, practitioners with a specific job to do, learners studying the space, competitor employees researching the market, analysts writing reports, or students working through coursework. We cannot tell from the query text alone. Throughout this report, findings describe queries in the pool. Any claim about who is doing the asking is labeled explicitly as inference from query shape, not measurement of audience.
Intent-inference layer. In addition to categorizing queries by topic, we bucket each query by the intent signal its structure carries. Four buckets:
These are structural inferences, not audience measurements. A learner-concept query could come from a founder learning finance terminology, a junior analyst, a student, or anyone else. What we can say with confidence is that the query shape is consistent with each intent, not that the person typing it belongs to any particular audience.
Limitations. Four to acknowledge.
(1) Impression volume is concentrated in a small number of high-traffic consumer queries; one query in the consumer pool ("how to make money online") alone accounts for 178,673 impressions, 71% of the total pool. We report both stratified per-segment and pooled numbers throughout.
(2) The B2B site pool includes one greenfield site with very low absolute volume (6 non-branded queries after filtering); its per-site numbers are directional only.
(3) The intent-inference heuristics are conservative and label many mid-length "how to" queries as ambiguous rather than force them into a bucket. Actual intent distribution is probably more nuanced than the four-bucket labeling suggests.
(4) The window is one calendar month; longitudinal comparability with Foundation #1's April 20 - May 20 window is approximate.
Impression share by intent-inference bucket. B2B site pool = 63.7% learner-intent, 1.2% purchase-consideration.
Queries with AI-native buyer-question shape exist in fintech AI search, but they're a small structural fraction. Only 8% of queries in the B2B fintech site pool show clear purchase-consideration signal (named vendor plus capability probe), and those queries account for just 1.2% of impressions.
The dominant intent in fintech AI search is learning, not buying. 63.7% of impressions on B2B fintech sites come from learner-intent queries: concept lookups like "what is embedded finance," "what is a fully burdened rate," "what is MRR."
Consumer fintech and B2B fintech behave very differently. Queries on consumer fintech sites (Site A, Site B) average 5.0 words impression-weighted and never end in a question mark. Queries on B2B fintech sites (Site C, Site D, Site E) average 5.7 words and 37% end in a question mark.
Foundation Study #1's 10-category taxonomy applies to fintech B2B queries with predictable shifts. Vendor comparison, enterprise readiness / compliance, and technical understanding are all proportionally larger than in the meta-AI-search-tool category. AI visibility tracking, Foundation #1's largest category at 28%, is completely absent.
The methodological correction: query shape ≠ audience identity. This study introduces an explicit intent-inference layer to separate what the data can and cannot support. Findings describe queries in the pool; where audience is inferred, the inference is labeled as inference from shape.
Foundation Study #1 documented a new grammar of AI-native queries: longer, more conversational, more comparative, more systems-oriented. That study's dataset was Atomic AGI's own site, a single-project, meta-AI-search-tool audience.
The obvious next question: does that grammar exist beyond meta-AI-search buyers? Is it a broad shift in how people talk to AI engines, or a segment-specific pattern that only shows up when the audience is already AI-native?
Study #2 tests that question in fintech, using a pool of five sites: two consumer-facing (Site A, Site B) and three that primarily serve B2B fintech buyers (Site C, Site D, Site E). The answer turns out to be more useful than a clean yes or no.
The AI-native buyer-question shape is present in the fintech B2B pool. It's structurally identical to what Foundation #1 documented. But it's rare, and it appears alongside a much larger population of queries with different shapes and different apparent intents. The interesting result is not "yes it exists" but "here's how much of the pool it actually is."
Concept lookups and how-to process queries account for 77% of impressions. Vendor comparison, integrations, and enterprise readiness each sit below 2%.
The volume imbalance is itself a finding: informational consumer queries on AI Overviews generate orders of magnitude more impressions than B2B buyer queries. Not because there are more consumer queries in the dataset (there are actually slightly fewer), but because each consumer query attracts far more impressions than each B2B buyer query.
A B2B query like "are there scalable banking compliance software platforms for growth?" clocks 26 impressions across a month; the consumer query "how to make money online" clocks 178,673.
Why? Two structural reasons that both suggest the same thing. First, consumer queries are shorter, more common, and generate more repeat searches by more people. Second, B2B queries in the buyer-question shape trigger AI Overviews that answer the question in place. Users read the AI-generated response and don't need to search again. A well-answered buyer query, by design, produces one impression, not many.
Consumer fintech pool (Site A + Site B, 185 queries, 246,284 impressions):
Consumer AI-search queries retain the classic Google-search grammar. Short, noun-heavy, command-shaped. "how to get p60," "how much is vat," "what is 1688," "how to make money online." No queries in the consumer pool end with a question mark.
Not one. Zero of the 185 queries are longer than 15 words. Whatever "AI-native language shift" is happening in the broader market, it hasn't reshaped these query streams.
B2B fintech site pool (Site C + Site D + Site E, 200 queries, 4,596 impressions):
Very different profile. The simple average query length is 37% higher than the consumer pool. Question marks appear on more than a third of queries. Long, multi-clause queries make up 12% of the pool. These are not the same query streams the consumer sites see.
Cleanest single discriminator: question-mark punctuation. Zero percent of consumer pool queries end with "?" versus 37% of B2B pool queries. That's a bright-line finding.
Consumer AI-search users still type queries the way they typed them into Google, as commands or fragments. B2B AI-search queries include a much larger share of full-sentence questions ending with a question mark. The grammar of the query itself carries a signal about the query stream it belongs to.
The four-bucket intent breakdown, applied to each segment:
Consumer segment:
B2B site pool:
The key finding surfaces here. In the B2B site pool, queries carrying purchase-consideration signal represent 8% of queries and 1.2% of impressions. Queries carrying learner-concept signal represent 26.5% of queries and 63.7% of impressions.
Read that again. Even in the pool most likely to contain B2B buyer queries, buyer-signal queries are 1.2% of impression traffic. The overwhelming majority of AI-search traffic entering these sites is from queries whose shape suggests someone learning a term ("what is a fully burdened rate," "what is embedded finance," "what is MRR"), not someone evaluating a purchase.
This is the single most important finding of Study #2, and it directly contradicts the reflexive framing many GEO/AEO discussions default to. A B2B fintech site winning AI-search impressions is not thereby winning buyer impressions - most of that traffic is learners.
One example of a clear purchase-consideration query (17 impressions, Site E): "does runway support scenario planning so finance teams can model different growth, hiring, and burn rate assumptions and share them with investors and board members?" Named vendor. Role mention. Multi-part capability probe. Specific use case. If any query in the pool suggests a professional evaluating a specific tool for a specific job, this one does.
One example of a clear learner-concept query (264 impressions, Site E): "what is a fully burdened rate." Foundational cost-accounting concept. Could be typed by a founder learning terminology, a junior finance analyst, an MBA student, a startup CEO writing a budget, or anyone else. Impossible to know from the text. What we can say: it's not a buyer-evaluation query. It's a definition lookup.
Applying Foundation Study #1's 10-category taxonomy to the B2B site pool produces a mostly clean mapping with three notable shifts. Category 1 (AI visibility, citations, and multi-engine tracking) has zero queries in the fintech pool - that was a meta-AI-search-tool-specific pattern that does not exist here.
Categories 7 (Vendor comparison), 8 (Enterprise readiness), and 9 (Technical / how it works) are all proportionally larger in the B2B fintech pool than they were in the Foundation #1 dataset.
Three fintech-specific residual categories emerge alongside the taxonomy: concept lookups, how-to process queries, and how-to calculation queries. Together these absorb the bulk of the B2B pool's impression volume, a direct reflection of the intent-inference finding that most impression traffic in this pool is learner-intent, not buyer-intent.
Categories below are ordered by impression share within the B2B site pool.
The recurring question shape:
"what is [fintech concept] / what does [term] mean / what are [technical construct]"
The biggest category in the B2B fintech pool by a wide margin. These are foundational-definition queries: "what is digital banking," "what is embedded finance," "what is MRR," "what is a fully burdened rate," "what does billings mean."
Structurally these queries are short (4-7 words typically), noun-based, and definitional. They carry no purchase-consideration signal, no requirements-listing, no vendor names. What they carry is a clear learner-intent signal: someone wants a definition, and AI Overviews delivers one directly.
What this says about the market. More than half of impression traffic entering B2B fintech sites via AI search is definitional. These queries do not indicate buyer intent by any reasonable inference, they indicate that people are learning terminology, and AI Overviews is answering them without a click.
The absolute impression volumes are non-trivial: "what is digital banking" alone captured 1,599 impressions on Site C in one month. But those impressions do not translate into purchase-consideration behaviour.
For marketers. Content that ranks for concept-definition queries generates category-authority signal, not pipeline signal. It's still worth doing, being the source AI Overviews cites when defining a category matters for long-term brand association, but it should be measured separately from buyer-facing content. Treating a rise in AI-search impressions on concept queries as "buyer traffic" would misread the pool.
The recurring question shape:
"how to [do a specific process or task in fintech]"
Short procedural queries: "how to grow deposits in banks," "how to improve customer experience in banking," "how to create investor updates," "how to score revenue ranges." Typically 5-9 words, imperative-flavoured, task-oriented.
The intent behind these queries is genuinely ambiguous. A junior analyst learning their job, a founder writing their first investor update, a consultant researching for a client project, and a mid-career professional refreshing on a specific task all plausibly type the same query. The query shape carries no discriminator between them.
What this says about the market. The how-to-process category sits between concept lookups (definitional, clearly learner-intent) and practitioner queries (multi-clause requirements, clearly professional-with-a-problem intent). It's the middle band, and its intent is genuinely mixed.
For marketers. Content in this category can be optimized for either learner audiences (evergreen tutorials) or practitioner audiences (advanced how-to with role-specific framing), and the query text won't tell you which is landing on your page. Rely on downstream signals, time on page, secondary engagement, downstream conversion, to distinguish.
The recurring question shape:
"are [X] safe / are [X] safer than [Y] / are [X] as secure as [traditional Y]"
Small category by query count but disproportionate impression share, driven mostly by two Site C queries: "are neobanks safe" (100 impressions) and "are digital banks safe" (33 impressions). Comparative queries positioning newer fintech categories against traditional banking incumbents.
The queries are comparative but not evaluative in the buyer sense. They're asking whether a broad category (neobanks, digital banks, mobile-only banks) is trustworthy relative to traditional banks. That's closer to a consumer-safety concern than a vendor-comparison inquiry.
What this says about the market. Even after several years of digital-banking growth, "are these things safe" remains a real category-validation question. Traditional-vs-digital comparisons continue to drive meaningful impression volume, the trust threshold has not been fully crossed for a segment of the audience.
For marketers. Trust-safety content around fintech category legitimacy still has readership. Neobank and digital-first fintech sites should assume their content will be read partly by people questioning whether the category itself is real, not just whether the specific brand is real.
The recurring question shape:
"are [product / category] safe / secure / legit"
Adjacent to C5 but at the specific-product level rather than the category level: "are online banks safe for savings," "are virtual banks safe," "are online-only banks secure enough for near-retirement savings."
Foundation #1's Category 3 (Trust, skepticism, and category validation) mapped to hype/legitimacy questions about AI-search tools. In fintech, this maps to safety/security questions about digital banking products. Same underlying category, different domain vocabulary.
What this says about the market. Digital-banking safety concerns persist as a real question in AI search. The queries are not sophisticated buyer-consideration questions; they're more foundational trust probes. This is the layer of the market that hasn't yet fully accepted the category.
For marketers. Evidence assets, regulatory disclosures, deposit-insurance details, security certifications, matter more here than product features. The buyer question isn't "which digital bank should I use" but "should I use a digital bank at all."
The recurring question shape:
"does [named vendor] provide / support / offer [specific capability]"
The clearest buyer-signal category in the pool. Named-vendor queries with capability probes: "does Runway support scenario planning so finance teams can model different growth, hiring, and burn rate assumptions?", "does Fondo provide real-time monitoring, full management capabilities, and strong support?", "does ncino provide a cloud-based banking operating system that automates loan origination?"
Structurally identical to Foundation #1's Category 7 pattern, the domain vocabulary changes (Runway, Fondo, Centime, ncino, Lumin Digital, Spendwise, Agent IQ in fintech; different vendor names in the meta-AI-search category) but the buyer-question shape is the same across both studies.
What this says about the market. The named-vendor-plus-capability shape is the most reliable structural marker of active-buyer intent in the data. When a query names a specific vendor and probes a specific capability, the person typing it is almost certainly evaluating that vendor for a specific job. This inference is stronger than any other intent inference we can make from query shape alone.
For marketers. Comparison content, head-to-head vendor pages, capability-comparison tables, "does [our tool] do X" landing pages, captures the highest-quality buyer-signal impressions in the pool. The volumes are small (single-digit impressions per query is normal) but the intent quality is the highest of any category. Optimizing here trades volume for signal purity.
The recurring question shape:
"how is AI used in banking / how are banks implementing [technology]"
Concentrated in the Site C dataset. Queries like "how ai is used in banking," "how are banks implementing ai-first operating models," "how are banks migrating to cloud-based core systems," "how are banks personalizing customer experience using ai."
Higher proportional share than Foundation #1's Category 9 (which was 0.5% of impressions in the meta-AI-search-tool category). Banking's technical complexity, regulatory constraints, legacy systems, core-banking migration, surfaces here as a distinct query stream.
What this says about the market. The banking-tech buyer segment includes people asking foundational "how does this work" questions at meaningful volume. This is closer to research or analyst intent than active-purchase intent.
For marketers. Technical-depth content in banking-tech (implementation guides, migration playbooks, architectural comparisons) captures a real audience segment that Foundation #1's dataset didn't show. Not a large audience, but a distinct one worth serving separately from product-marketing content.
The recurring question shape:
"are there AI-powered / automated / ML-driven tools that [specific capability]"
The fintech B2B version of Foundation #1's Category 4. Queries about automation and AI-driven capabilities: "are there AI-driven platforms that integrate with existing banking systems?", "are there transaction data enrichment APIs with machine-learning features?", "how accurate is AI-powered cash flow forecasting for a mid-market company with several entities?"
Multi-clause, requirements-oriented, systems-aware. Same buyer-question shape as Foundation #1 documented, applied to fintech automation categories.
What this says about the market. Buyers in the fintech B2B pool are actively looking for AI-augmented tooling, but with skepticism about accuracy. The queries are not "does AI exist here" but "does the AI here actually work for my scale and my problem."
For marketers. Content addressing AI accuracy at specific scales (small business, mid-market, enterprise), and specifically for use cases with clear success criteria (forecasting, anomaly detection, categorization), maps cleanly to the queries in this category.
The recurring question shape:
"how much does [service] cost / pricing for [category]"
Cost-probing queries: "how much does a fractional cfo cost for startups," "how much do fp&a services cost for startups," "how much venture debt can a saas company raise."
Foundation #1's Category 6 was about free/affordable AI-tool tiers. In fintech B2B, the pricing category is more about service cost benchmarks, what a fractional CFO or FP&A engagement typically costs, or how much a company can raise. Same category concept, different domain application.
What this says about the market. Fintech buyers include cost-transparency as a live question. Even in a category with historically opaque pricing (fractional CFO services, professional financial advisory), buyers are typing direct cost questions into AI search.
For marketers. Publishing transparent pricing bands, even ranges, captures this segment. Sites that require "contact us for pricing" lose visibility on these queries. Even a "typical range: $5K–$15K/month for pre-Series-B, $15K–$40K/month post-Series-B" rough band would be indexable content.
The recurring question shape:
"does [product] integrate with [ERP / QuickBooks / accounting system]"
Fintech B2B integration queries: "are there tools that help unify data from erp, banks, and payroll into one cash view?", "how do i find an accounting partner experienced with quickbooks cleanup, payroll management, and monthly financial dashboards?"
Foundation #1's Category 2 was about AI-search-tool integrations with marketing analytics (GA4, Looker, Slack). In fintech, integration means ERP, accounting software, payroll systems, banking core systems, the financial operations stack. Same category structure, entirely different vocabulary layer.
What this says about the market. Fintech buyers evaluate tools by what they connect to. A cash-management tool that doesn't plug into the buyer's existing accounting system faces a structural disadvantage that shows up in the query language itself.
For marketers. Integration lists on product pages should name specific systems (QuickBooks, Xero, NetSuite, SAP, specific banking platforms) rather than generic categories ("your accounting software"). AI-search-visible integration pages get captured on the specific-vendor queries in this category.
The recurring question shape:
"are there [scalable / compliant / secure] platforms for [enterprise use case]"
Regulatory and governance-oriented queries: "are there scalable banking compliance software platforms for growth?", "what are the best solutions for customer due diligence and kyc onboarding?", "who offers a dependable ongoing due diligence solution with strong automation capabilities?"
Proportionally larger than Foundation #1's Category 8 (which was 1.2% of impressions in meta-AI-search). Banking is compliance-heavy, and the queries reflect that: KYC, KYB, AML, due diligence, banking compliance software all show up as distinct query streams.
What this says about the market. Compliance and enterprise-readiness are not incidental concerns in fintech B2B - they're a distinct buyer-consideration category with real impression volume. Foundation #1 was in a category that hadn't yet reached this stage of maturity; fintech is well past it.
For marketers. Compliance content, regulatory certifications, and enterprise-governance features belong prominently on marketing pages, not buried in security documentation. The queries in this category are typed by people who need to satisfy internal procurement or regulatory checks before purchase.
The recurring question shape:
"how to calculate [financial metric] / how to forecast [metric]"
Formula-oriented queries: "how to calculate fully burdened labor rate," "how to calculate cash runway," "how to forecast mrr and arr," "how to calculate unit economics saas."
A fintech-specific residual category. Foundation #1 had no equivalent, meta-AI-search-tool buyers weren't asking "how do I calculate X." Fintech, with its metric-heavy vocabulary (ARR, MRR, LTV, CAC, unit economics, burden rates, runway), sees a distinct stream of calculation-methodology queries.
What this says about the market. People typing calculation queries into AI search are looking for methodology, not tools. The queries are indistinguishable from what someone would type into Google, short, procedural, non-comparative. Fintech tools appear in the results, but the query itself is educational.
For marketers. Calculator content and methodology-explainer pages capture this segment. Fintech tool vendors that publish accessible calculators or step-by-step methodology guides pick up these impressions even though the query intent is learning, not buying.
The recurring question shape:
"how does [X] affect [business outcome]"
Effectively empty in the fintech B2B pool. One query in the entire dataset, "how does venture debt affect valuation" (1 impression), carries anything close to a business-impact signal, and even that is really a learner-concept query dressed as an impact question.
Foundation #1's Category 10 was also small (0.5% of impressions). The pattern holds: attribution and business-impact queries are not what people type into AI search at meaningful volume, in either the meta-AI-search category or in fintech.
What this says about the market. Attribution is a marketer-side and CFO-side question, not a buyer-side question. Neither vertical shows it as a live query stream.
This is consistent with the LinkedIn comment on Foundation #1's launch: attribution is structurally a measurement problem posed by the party trying to measure, not a discovery question posed by the party doing the buying.
For marketers. Same read as Foundation #1: attribution-proof content matches internal marketing-team debates more than external buyer queries. Continue producing it for sales and internal reporting; don't expect it to drive AI-search impressions.
The largest category in Foundation Study #1 (28% of impressions) does not exist in the fintech B2B pool. Zero queries in this dataset ask about tracking brand visibility across LLMs, monitoring citations, or measuring AI mentions.
That absence is itself a finding. The AI-visibility-tracking category in Foundation #1 was populated by an audience that was already asking about AI search as a category: people evaluating AI-search analytics tools, curious about how to measure their own presence, or trying to understand how to appear inside AI answers.
That audience does not naturally appear on fintech sites. Fintech buyers aren't asking Site C how to track their brand mentions across ChatGPT, that's not what Site C is for, and it's not what fintech buyers are worried about.
The finding this exposes is broader than fintech: Foundation #1's biggest category was likely disproportionately large because the dataset was a meta-AI-search-tool site, so the audience was pre-selected for AI-visibility interest. Extending Foundation #1's category taxonomy to non-meta verticals shows that the top category doesn't generalize. It was a category-specific artefact, not a universal buyer concern.
The two-segment split confirms and refines Foundation Study #1's language-shift thesis. The AI-native buyer-question grammar exists in fintech, and its structural signatures are consistent with what Foundation #1 documented in the meta-AI-search-tool category. It appears in queries with named vendors, multi-clause requirements, role mentions, and question-mark punctuation.
But it is not the dominant grammar of fintech AI-search. Most impression traffic in this pool, both consumer and B2B, comes from queries with older, shorter, informational grammar. The AI-native buyer shape is a real structural population, but it is a small one.
The consumer segment shows the language shift essentially absent. Not one query in 185 uses question-mark punctuation. Not one is longer than 15 words. Whatever AI-native grammar is developing in the broader market, it has not reshaped how consumer fintech users type into AI-search interfaces.
The B2B site pool shows the language shift present but concentrated. "37% of queries end with "?" 12% run longer than 15 words". But when we look at intent inference, only 8% of these queries carry clear purchase-consideration signal. Most of the shift is happening in queries whose intent is learning or practitioner-with-a-problem, not buying.
Three structural readings follow.
The AI-native grammar exists but is layered on top of a much larger informational pool. Marketers reading Foundation #1 and expecting the buyer segment to dominate their AI-search traffic will be disappointed. In fintech at least, the buyer segment is a narrow band.
Purchase-consideration queries appear where you'd expect them and are shaped consistently across verticals. Named vendor + capability probe queries in Site E (SaaS-CFO tooling) and Site C (banking-tech) share the same grammatical structure as similar queries in Foundation #1's meta-AI-search-tool pool. The buyer question shape generalizes; only the topical vocabulary changes.
AI Overviews may be actively suppressing buyer-signal impression volume. The clearest buyer-signal queries have very low impressions, often single-digit, because AI Overviews synthesizes an answer directly from the site's content. The user gets what they need and doesn't need to keep searching.
This is a byproduct of doing GEO/AEO well: successful capture means fewer repeat impressions. Any measurement of buyer intent that leans purely on impression volume will systematically undercount the buyer segment.
Framing note first. Findings in this study describe queries in the analyzed pool. They do not directly measure buyer behavior. If you build strategy from a query-pattern finding, you're building from what buyers plausibly typed, not from what buyers definitely did. The distinction matters because it changes what you can claim about outcomes.
Three concrete actions for marketing teams.
1. Do not confuse AI-search traffic volume with buyer-signal volume. Most impression traffic in fintech AI-search is learner-intent. Winning impressions on "what is embedded finance" is a category-authority signal, not a pipeline signal. Marketing teams that measure AI-search success purely by impressions will misread their actual buyer capture.
2. If you are targeting the buyer segment specifically, look for the query-shape markers. Named vendor plus capability probe. Multi-clause requirements with role mentions. Question-mark punctuation on long queries. These are the structural signals of buyer-consideration intent in the AI-search data. Content that captures these queries specifically is doing different work than content that captures learner-intent queries.
3. Segment your AI-search analytics by intent, not just by page or by keyword. The dominant categorization tools currently treat all AI-search traffic as equivalent. This study shows those queries are not equivalent - they represent very different apparent intents with very different implications for pipeline. Manual intent segmentation is currently the only reliable way to see what portion of your AI-search capture actually looks like buyer traffic.
This is the second study in our monthly research series on AI-native search behaviour. Next month (August 2026): a longitudinal update on Foundation Study #1, extending the analysis window from 30 to 60 days to test whether the category shares documented in June have held, shifted, or grown.
September: the same impression-weighted linguistic-pattern methodology applied to a B2B SaaS project pool. October: cross-vertical quarterly synthesis bringing the fintech, longitudinal, and SaaS results together.
Subscribe at omnius.so/research/subscribe to get each issue when it ships.
For citation: Omnius Research, "Fintech AI Search: Mostly Learning, Sometimes Buying," July [DATE], 2026. Methodology extends Omnius Research, "What People Ask AI Search About AI Search" (June 22, 2026), and follows Druck & Smith, "Demystifying Randomness in AI" (Graphite, 2026).
Dataset: Atomic AGI Google Search Console AI Mode and AI Overview queries, June 1 – 30, 2026, from five fintech projects (Site A, Site B, Site C, Site D, Site E). 385 non-branded queries, 250,880 impressions, two segments, four intent buckets.
Learn how to scale user acquisition without scaling costs from our findings. We spent years exploring, so you don't have to.
Omnius is a B2B SEO & GEO agency; partnering up exclusively with SaaS, Fintech & AI companies. The result? Compounding growth made through organic positioning everywhere people search for information, including both Google & LLM search engines.
