Independent research on how AI answer engines shortlist B2B softwareUnited States

Does AI put you on the buying shortlist?

Cited & Selected studies whether AI answer engines such as ChatGPT mention, cite and recommend B2B software suppliers when US buyers ask buying questions. For marketing, SEO and product-marketing leaders: read the research, run the free example tools, and compare monitoring providers—no sign-up needed.

Opens with example data. Email optional.Read the research

IndependenceNo supplier can pay for placement or ranking.
SeparationMentions, citations and recommendations are measured apart.
DenominatorsMissing answers stay missing—never counted as zero.

Which customer-support platforms suit a US B2B software team that needs Salesforce integration?

Scope preview — collection not started

This check would ask 6 buying prompts on 2 supported surfaces, with 3 fresh-session observations per prompt . Each answer is scored three ways, never blended into one score .

6buying prompts

2×3surfaces × repeats

36planned observations

coverage = prompts with ≥1 recommendation ÷ prompts successfully tested

Not collected

Fictional brand	Mentioned	Recommended
HarbourDesk	Demo data	Demo data
RelaySupport	Demo data	Demo data
CaseBridge	Demo data	Demo data
QueuePilot	Demo data	Demo data

Shortlist outcomes appear only after collection. An empty cell is not a zero.

Fig. 1 — Anatomy of a shortlist check: the unbranded buying question, a one-pass conclusion, a visible denominator and attached sources. Select a numbered citation to open its source.

Covers: Whether your company is mentioned, cited or recommended when buyers ask ChatGPT, Google AI Overviews and AI Mode, Copilot and Perplexity—and what to buy, build or fix about it.
Helps: Marketing, growth, SEO, content and communications leaders at US B2B software companies—plus the founders, analysts and procurement teams who have to defend the decision.
Why it's useful: Buyer-side and independent. Every number shows its denominator, every claim shows its source, and the answer can be “use what you already have”. No supplier pays for placement.

Prompt set

The buyer asks“Are we even testing the questions our buyers ask?”

Choose the questions that could change a shortlist.

Start with customer questions, requirements and comparisons. Inferred prompts are labelled as inferred. Edit the four fields and the log rewrites itself. Prefilled with fictional example inputs.

Being named is not the same as being recommended.

The buyer asks“Are rivals recommended where we're only named?”

One disclosed population. Three different outcomes. Missing answers remain missing.

Fig. 3 — How every rate on this desk is built.

Not collected

The matrix below is empty because no answers have been collected. Load the fictional demo to see how the measures behave.

The sampling, scoring and failure-handling rules behind every number on this page.

Inspect the answer before explaining the gap.

The buyer asks“Which pages is the answer actually drawing on?”

The answer, the cited page and the collection conditions behind every finding.

Fictional demo evidence surface: Surface A, consumer interface session: fs-0417-b demo record, not a real timestamp

Exact prompt

“Which customer-support platforms suit a US B2B software team that needs Salesforce integration and audit-ready security documentation?”

For a US B2B software team with a Salesforce-first stack, RelaySupport and CaseBridge are the options most often put forward. RelaySupport documents a native Salesforce connector and publishes a SOC 2 report on request .

CaseBridge is frequently compared on audit logging, and an independent comparison lists it alongside QueuePilot for regulated buyers . Community threads describe HarbourDesk's integration as workable but limited to case sync .

If security documentation is the deciding factor, review each vendor's trust centre directly .

Three different relationships, never one badge

Linked by the assistantMentioned on the cited pageRecommended in the answer

Paste your own answer transcripts and map the domains cited beside each competitor.

Tool, build or expert

The buyer asks“What should we actually buy—if anything?”

Buy the missing capability—not another GEO label.

Four honest routes, including no additional purchase. Nothing is preselected.

What capability is missing?

Select a route to see what it involves, what to check and what you still need.

A dozen quick questions, transparent scores for all four routes and a 30/60/90-day brief.

Start with six buying questions, not six months of retainer.

The buyer asks“Can we get something useful before talking to anyone?”

Useful exports, clear limits and no compulsory sales call. Each tool runs in your browser on what you enter.

Two pages from the buyer's guide: choosing prompts buyers really ask, and a table comparing tool, build, expert and existing-team routes.

Free guide

The AI Search Visibility Buyer's Guide

Checklists, route criteria, USD planning bands and RFP questions you can take into your next supplier conversation.

Mentions, citations and recommendations
Choosing prompts buyers really ask
Tool, build or expert: the criteria
Planning budgets in USD
RFP questions suppliers should answer

A monitoring platform and a PR programme are different purchases.

The buyer asks“Who measures, and who actually does the work?”

Compare measurement, implementation and validation separately. Alphabetical; sponsorship never changes the order.

Your comparison

Real companies, listed by name and category for identification only. Nothing here has been assessed, ranked or endorsed, and no commercial relationship or partnership exists with any of them. Unknown values read “Not assessed”.

Research you can inspect. Decisions you can explain.

The buyer asks“How do I explain this decision to the board?”

Open the research library

Twelve articles, the research programme, coverage map and collection protocol.

A mention is not a recommendation. A citation is not a conversion.

Independence

Any research fees or referral income will be recorded separately and disclosed. Suppliers cannot buy placement, ratings or sort order.

Sampling

Prompts, surfaces, repetitions and access routes are disclosed first. API and consumer-interface observations are labelled differently.

Scoring

Every rate shows its numerator and denominator. Failed observations are listed and excluded, never counted as absence.

Corrections

Methodology changes are dated beside affected results. Market figures are directional planning estimates, not verified statistics.

These rules are the published collection protocol. No answer data has been collected or published yet. Read the full collection protocol

Start here

Find out if exclusion is real before you fund a fix.

Six prompts, two surfaces, three repeats: a scope plan and evidence template, shown on this page. Email optional.