Chatbots and conversational AI for support
Asked as “customer support chatbot platform”, and as “conversational AI tool for customer service”, on behalf of a mid-market B2B software company. 19 first choices recorded across the direct, paraphrase, budget and scale prompts, six models each.
Standing · first-choice share
42%
Clear leader
42Intercom Fin AI Agent16Fin AI Agent16tawk.to26others
42% of first choices, clear leader.
The standing
Share is the count of first choices across the direct, paraphrase, budget and scale prompts, over all six models, for a mid-market B2B software company. Ordered by share.
ProductFirst-choice shareNegative rateLabelsQuadrantSince September 2026
01
Intercom Fin AI Agent42%21%24endorsed leadernewNew since September 2026: not ranked then, 42% now.
Intercom Fin AI Agent42%21%24endorsed leadernewNew since September 2026: not ranked then, 42% now.Show the two products at 0%, ordered by negative rate
04
Zendesk AI Agents0%30%10criticized challengernewNew since September 2026: not ranked then, 0% now.
Zendesk AI Agents0%30%10criticized challengernewNew since September 2026: not ranked then, 0% now.The floor is 12 points of share, measured: how far the models move a leader on their own when the same questions are asked twice with nothing changed. A larger change is movement; a smaller one can't be told apart from repeat variation, and both are shown. Each index measures its own floor from its own repeat, over its own categories, so the floor differs from one index to the next. How the floor is measured
Bars are the share of first choices, 0 to 100Every product with at least 10 labels here. Every product name links to its product page.
Head to head
Recommended versus criticized
Every product with at least 10 labels here, on both axes. The 30% line names a quadrant, not the verdict above: that one needs more than 40%.
Criticized challengerCriticized default
Negative label rate →
01
02
0304Accepted challengerEndorsed leader
0%First-choice share → · lines at 30% share and 25% negative60%
Key
01Intercom Fin AI Agent42%
Intercom Fin AI Agent42%02Zendesk AI5%
Zendesk AI5%03Ada0%
Ada0%04Zendesk AI Agents0%
Zendesk AI Agents0%What they warned about
Four of six models held their first choice under the paraphrase. GPT-5.6 Sol and Perplexity Sonar Pro changed. A high negative share on a product with few labels is a warning. A low share on a product with many labels is salience, not sentiment.
Intercom Fin AI Agent21%
5 of 24 labels negative · 5 of 6 models
“Intercom has one of the best AI agents on the market, but its $0.99-per-resolution pricing model ... is heavily criticized by small-to-medium businesses.” Gemini 3.1 Pro, negative prompt
Zendesk AI Agents30%
3 of 10 labels negative · 3 of 6 models
“It is incredibly powerful but often requires a steeper learning curve and a higher total cost of ownership than simpler tools.” Gemini 3.1 Pro, comparative prompt
Ada22%
4 of 18 labels negative · 2 of 6 models
“Ada is the clearest named example I found: G2 reviewers consistently flag implementation as "a huge, time-consuming project," making it better suited to large CX programs than smaller teams” Claude Opus 5, negative prompt
Chatbase50%
2 of 4 labels negative · 2 of 6 models
“they are less clearly positioned as full customer-support suites than Tidio” Perplexity Sonar Pro, budget prompt
What they cite
Six of the six models returned a source list in this edition: six return the pages they cited.
Sites the answers cite
32 of 36 answers in this category came back with a source list, from 6 of 6 models: citations where the model returns them, or the search results it consulted. 551 links across 243 sites, every framing counted. Ranked by the number of answers carrying the site or page. None of the 36 answers across every segment cited this index's own page for the category.
vendor site · Intercom14 answers · 29 citations · 4 models
vendor site · Featurebase11 answers · 19 citations · 4 models
11 answers · 18 citations · 3 models
vendor site · Fin.ai10 answers · 21 citations · 4 models
vendor site · Zendesk9 answers · 10 citations · 3 models
vendor site · My AskAI7 answers · 14 citations · 4 models
7 answers · 11 citations · 3 models
vendor site · Tidio7 answers · 9 citations · 6 models
vendor site · Capterra6 answers · 8 citations · 1 model
vendor site · Zendesk6 answers · 7 citations · 3 models
vendor site · Enjo6 answers · 6 citations · 4 models
5 answers · 10 citations · 3 models
Pages the answers cite
The ten pages named in the most answers, by full address. A page here is one the models returned with a recommendation, not one the index endorses.
8 answers · 3 models
7 answers · 3 models
7 answers · 3 models
7 answers · 3 models
5 answers · 1 model
5 answers · 4 models
5 answers · 3 models
5 answers · 4 models
4 answers · 2 models
4 answers · 3 models
Search against answers
Each company's standing in the answers beside its site's footprint in Google search, one row a site: the products the models named on it with their shares, and the share they add up to; monthly searches on Google, and DataForSEO's estimate of AI search demand (modeled from search signals, directional, not a count of queries to any assistant), for the most-searched of the company's and its products' names (the name is in each row's hover text); estimated monthly organic visits to the site; and its best position in Google's top ten for “best customer support chatbot platform”, “customer support chatbot platform”. US estimates from DataForSEO and Google's Ads Transparency Center. A small company's site, or a mid-sized company's site for its flagship, is marked company; a product on a large parent's site (Google, Microsoft) has no site figures. A column with no figures for this category is left out, and an empty cell means none were seen, not none exist. Two measurements side by side: neither is read as the cause of the other.
| Company and products | Share | Own site cited | Name searches, Google | AI search demand, est. | Organic visits | Paid search a month, est. | Google ads, last 30 days | Best Google position |
|---|---|---|---|---|---|---|---|---|
| Intercom Fin AI Agent intercom.com | 42% | 159 | 210 | 3 | 64,348 | $3,304 | 120 | |
| tawk.to tawk.to | 16% | 41 | 2,900 | 63 | 13,444 | 7 | ||
| Fin AI Agent | 16% | |||||||
| Tidio tidio.com | 11% | 236 | 6,600 | 102 | 79,128 | 64 | ||
| Zendesk AI zendesk.com Zendesk AI 5%, Zendesk AI Agents 0% | 5% | 102 | 880 | 48 | 5,365,445 | $16,181 | 63 | 1 |
| Helply helply.com | 5% | 5 | 210 | 7 | 9,626 | 0 | ||
| HubSpot parent site HubSpot Chatbot 5% | 5% | 11 | 368,000 | 9,776 | ||||
| Ada ada.cx | 0% | 50 | 201,000 | 50,587 | 23,398 | 45 | 8 | |
| Salesforce parent site Salesforce Agentforce for Service 0%, Salesforce Service Cloud 0% | 0% | 223 | 550,000 | 60,632 | ||||
| Crisp crisp.chat | 0% | 86 | 60,500 | 36,276 | 29,685 | 2 |
Google's top ten for the two searches hold 17 sites; 3 of them are among the sites the models cited here (eesel.ai, thecxlead.com, zendesk.com). The first result for “best customer support chatbot platform” is twig.so.
How they answered
Six framings of the same buying question, each sent to every model in a fresh session with search on. One row per model, so a row shows whether it held its answer under rewording, what it named when cost was the constraint, and what it argued against. Computed from the raw judge labels.ShowHide
How they answered
Six framings of the same buying question, each sent to every model in a fresh session with search on. One row per model, so a row shows whether it held its answer under rewording, what it named when cost was the constraint, and what it argued against. Computed from the raw judge labels.
ShowHideModelDirect“What is the best customer support chatbot platform for a mid-market B2B software company?”Paraphrase“Which conversational AI tool for customer service would you recommend to a mid-sized B2B software company?”Comparative“What are the top customer support chatbot platforms and how do they differ?”Budget-constrained“What is the best customer support chatbot platform for a company with a limited budget?”Scale-constrained“We are a 500 person company evaluating a customer support chatbot platform. What should we look at?”Negative“Which customer support chatbot platforms should I avoid or be cautious about?”
Claude Opus 5Intercom Fin AI Agentagainst: AdaIntercom Fin AI AgentHeldagainst: Decagon, Freddy, Salesforce Service Cloud, SierraIntercom Fin AI AgentTidioagainst: Ada, Intercom Fin AI Agent, Sierrano first choiceagainst: Ada
Three alternatives
Plain, Pylon, Zendesk AI AgentsOne alternative
Zendesk AI AgentsSix alternatives
Ada, Decagon, Freshworks Freddy AI Agent, Salesforce Agentforce for Service, Sierra, Zendesk AISix alternatives
Chatfuel, Chatwoot, Crisp, Freshdesk, ManyChat, Zoho SalesIQClaude Opus 4.8Intercom Fin AI AgentIntercom Fin AI AgentHeldIntercom Fin AI AgentHubSpot Chatbotagainst: Chatbase, Intercom Fin AI Agent, Lindy, eesel AIno first choicenothing named
Two alternatives
Freshworks Freddy AI Agent, Zendesk AIThree alternatives
Ada, Freshworks Freddy AI Agent, Zendesk AIFive alternatives
Ada, Gorgias AI Agent, Salesforce Agentforce for Service, Tidio, Zendesk AI AgentsTwo alternatives
ChatBot.com, TidioGPT-6 AstraFin AI AgentFin AI AgentHeldno first choicetawk.toZendesk AIagainst: Freshworks Freddy AI Agent, Gorgias AI Agent, Intercom Fin AI Agent, Tidio Lyro, Zendesk AI Agents
One alternative
Zendesk AI AgentsTwo alternatives
Ada, Zendesk AI AgentsTwo alternatives
Chatwoot, TidioTwo alternatives
Intercom Fin AI Agent, Salesforce Agentforce for ServiceGPT-5.6 SolIntercom Fin AI Agentagainst: AdaFin AI AgentChangedIntercom Fin AI Agent, Zendesk AI Agentstawk.toagainst: HubSpot Service Hubno first choiceagainst: Drift, Freshworks Freddy AI Agent, Gorgias AI Agent, Intercom Fin AI Agent, Salesforce Agentforce for Service, Tidio Lyro, Zendesk AI Agents
Two alternatives
HubSpot Service Hub, Zendesk AIThree alternatives
Freshdesk, Salesforce Service Cloud, Zendesk AISeven alternatives
Ada, Freshworks Freddy AI Agent, Genesys Cloud Agentic Virtual Agent, Gorgias AI Agent, HubSpot Breeze Customer Agent, Microsoft Copilot Studio, Salesforce Service CloudThree alternatives
Crisp, Freshchat, TidioGemini 3.1 ProIntercom Fin AI AgentIntercom Fin AI AgentHeldIntercom Fin AI Agentagainst: Zendesk AI Agentstawk.toagainst: HubSpot CRMno first choiceagainst: Ada CX, Chatfuel, Intercom Fin AI Agent, ManyChat
Four alternatives
Forethought, My AskAI, TeamSupport, Zendesk AIFour alternatives
Ada, DevRev, Forethought, Zendesk AIEight alternatives
Ada, Botpress, Chatarmin, Decagon, Drift, Gorgias AI Agent, Halo AI, TidioThree alternatives
Chatbase, Crisp.chat, TidioPerplexity Sonar ProIntercom Fin AI AgentHelplyChangedZendesk AIagainst: Amazon Lex, Help Scout, RasaTidioagainst: Chatbase, Freshdesk, Quidgetno first choiceagainst: Cognigy, Google Dialogflow CX, Kore.ai, Rasa, Sobot AI Agents
Four alternatives
Ada, Plain, TeamSupport, Zendesk AIThree alternatives
Intercom Fin AI Agent, Zendesk AI, eesel AIFive alternatives
Ada, Gorgias AI Agent, Helpshift, Intercom Fin AI Agent, TidioTwo alternatives
Crisp, tawk.toBold is the first choiceAlternatives are counted; the count opens them.What the answer argued against
The record
One row per call: the version string exactly as returned, whether the model searched, sources cited, and latency.
Thirty-six rows: every prompt, every model, every answer.
| Prompt | Model | Version string | Time (UTC) | Searched | Sources | Latency |
|---|---|---|---|---|---|---|
| Direct recommendation | Claude Opus 5 | claude-opus-5 | 2026-09-09 19:28 | yes | 34 | 56 s |
| Direct recommendation | Claude Opus 4.8 | claude-opus-4-8 | 2026-09-09 19:30 | yes | 28 | 58 s |
| Direct recommendation | GPT-6 Astra | gpt-6-astra | 2026-09-09 19:32 | yes | 7 | 35 s |
| Direct recommendation | GPT-5.6 Sol | gpt-5.6-sol | 2026-09-09 19:33 | yes | 7 | 38 s |
| Direct recommendation | Gemini 3.1 Pro | gemini-3.1-pro-preview | 2026-09-10 17:04 | yes | 7 | 33 s |
| Direct recommendation | Perplexity Sonar Pro | sonar-pro | 2026-09-09 19:30 | yes | 20 | 5 s |
| Paraphrase | Claude Opus 5 | claude-opus-5 | 2026-09-09 19:36 | yes | 24 | 59 s |
| Paraphrase | Claude Opus 4.8 | claude-opus-4-8 | 2026-09-09 19:37 | no | 0 | 18 s |
| Paraphrase | GPT-6 Astra | gpt-6-astra | 2026-09-09 19:39 | yes | 6 | 30 s |
| Paraphrase | GPT-5.6 Sol | gpt-5.6-sol | 2026-09-09 19:40 | yes | 9 | 67 s |
| Paraphrase | Gemini 3.1 Pro | gemini-3.1-pro-preview | 2026-09-10 17:06 | no | 0 | 26 s |
| Paraphrase | Perplexity Sonar Pro | sonar-pro | 2026-09-09 19:38 | yes | 20 | 6 s |
| Comparative | Claude Opus 5 | claude-opus-5 | 2026-09-09 19:45 | yes | 38 | 76 s |
| Comparative | Claude Opus 4.8 | claude-opus-4-8 | 2026-09-09 19:50 | yes | 32 | 148 s |
| Comparative | GPT-6 Astra | gpt-6-astra | 2026-09-09 19:53 | yes | 8 | 45 s |
| Comparative | GPT-5.6 Sol | gpt-5.6-sol | 2026-09-09 19:55 | yes | 10 | 44 s |
| Comparative | Gemini 3.1 Pro | gemini-3.1-pro-preview | 2026-09-10 17:07 | yes | 9 | 42 s |
| Comparative | Perplexity Sonar Pro | sonar-pro | 2026-09-09 19:51 | yes | 20 | 10 s |
| Budget constrained | Claude Opus 5 | claude-opus-5 | 2026-09-09 19:57 | yes | 38 | 50 s |
| Budget constrained | Claude Opus 4.8 | claude-opus-4-8 | 2026-09-09 19:58 | yes | 36 | 47 s |
| Budget constrained | GPT-6 Astra | gpt-6-astra | 2026-09-09 20:00 | yes | 4 | 21 s |
| Budget constrained | GPT-5.6 Sol | gpt-5.6-sol | 2026-09-09 20:01 | yes | 3 | 22 s |
| Budget constrained | Gemini 3.1 Pro | gemini-3.1-pro-preview | 2026-09-10 17:09 | yes | 10 | 40 s |
| Budget constrained | Perplexity Sonar Pro | sonar-pro | 2026-09-09 19:59 | yes | 20 | 5 s |
| Scale constrained | Claude Opus 5 | claude-opus-5 | 2026-09-09 20:03 | yes | 16 | 56 s |
| Scale constrained | Claude Opus 4.8 | claude-opus-4-8 | 2026-09-09 20:04 | no | 0 | 21 s |
| Scale constrained | GPT-6 Astra | gpt-6-astra | 2026-09-09 20:06 | yes | 5 | 39 s |
| Scale constrained | GPT-5.6 Sol | gpt-5.6-sol | 2026-09-09 20:07 | yes | 5 | 49 s |
| Scale constrained | Gemini 3.1 Pro | gemini-3.1-pro-preview | 2026-09-10 17:10 | no | 0 | 24 s |
| Scale constrained | Perplexity Sonar Pro | sonar-pro | 2026-09-09 20:04 | yes | 20 | 8 s |
| Negative framing | Claude Opus 5 | claude-opus-5 | 2026-09-09 20:10 | yes | 39 | 74 s |
| Negative framing | Claude Opus 4.8 | claude-opus-4-8 | 2026-09-09 20:12 | yes | 29 | 64 s |
| Negative framing | GPT-6 Astra | gpt-6-astra | 2026-09-09 20:15 | yes | 5 | 38 s |
| Negative framing | GPT-5.6 Sol | gpt-5.6-sol | 2026-09-09 20:18 | yes | 7 | 97 s |
| Negative framing | Gemini 3.1 Pro | gemini-3.1-pro-preview | 2026-09-10 17:11 | yes | 15 | 37 s |
| Negative framing | Perplexity Sonar Pro | sonar-pro | 2026-09-09 20:13 | yes | 20 | 9 s |
Noise floor in this category
Flips between the edition run and its calibration repeat. Six prompts per model is a small sample; the index-wide floor is the number to trust.
ShowHide
Claude Opus 5
2 of 4 flipped
Claude Opus 4.8
0 of 4 flipped
GPT-6 Astra
3 of 4 flipped
GPT-5.6 Sol
3 of 4 flipped
Gemini 3.1 Pro
2 of 4 flipped
Perplexity Sonar Pro
2 of 4 flipped
Normalization in this category
Every judgment call made between the raw labels and the numbers above, listed so it is visible and reversible.
ShowHide
Category-scoped readings
Fin read as Intercom Fin AI Agent
Fin (formerly Intercom) read as Intercom Fin AI Agent
Fin by Intercom read as Intercom Fin AI Agent
Freshworks (Freddy AI) read as Freshworks Freddy AI Agent
Gorgias read as Gorgias AI Agent
HubSpot read as HubSpot Service Hub
HubSpot Chatbot Builder read as HubSpot Chatbot
HubSpot Customer Agent read as HubSpot Breeze Customer Agent
Intercom read as Intercom Fin AI Agent
Intercom (Fin AI Agent) read as Intercom Fin AI Agent
Intercom (Fin AI) read as Intercom Fin AI Agent
Intercom (with Fin AI) read as Intercom Fin AI Agent
Intercom Fin read as Intercom Fin AI Agent
Salesforce read as Salesforce Agentforce for Service
Salesforce Agentforce read as Salesforce Agentforce for Service
Sobot read as Sobot AI Agents
Zendesk read as Zendesk AI
Zendesk (Zendesk AI / AI Agents) read as Zendesk AI
Zendesk (Zendesk AI / Advanced Bots) read as Zendesk AI
Zendesk (with AI agents) read as Zendesk AI
Unresolved, counted raw
None awaiting human review.
Discontinued, still offered
No shut-down product was recommended here.