Who Do AI Engines Recommend for Conversational AI Consulting?
By Christi AkinwumiSeptember 12, 2026In this article
Most advice about getting cited by AI engines is opinion. Nobody publishes the score. So ICX built a scoreboard, ran it, and is publishing the first edition here, including the parts where ICX loses.
Below is the first edition of the ICX AI Citation Index, collected on September 12, 2026. It answers one question a buyer would actually ask: when someone types a conversational AI consulting question into an AI engine, who gets named?
What is the ICX AI Citation Index?
The index is a fixed set of 28 prompts, run through live AI engines on a schedule, and scored on one thing: does a given brand get named or cited in the answer?
The 28 prompts split into two groups. Twenty-two are category prompts. They are open questions with no brand in them, such as “Who are the best conversational AI consultants for enterprise customer service?” These are the prompts where discovery is won or lost. Six are branded prompts that name ICX or its founder. Those test whether an engine describes the firm correctly once it already knows the name.
The prompt set was written in June 2026 as a tracking list and has not changed since. It covers the buyer identity question, SaaS support cost, service-specific hiring, ROI, Dallas-Fort Worth local search, and topical authority. The full list is in the published data file linked below.
ICX has explained the ideas behind this before, in what AEO and GEO actually mean and in how Google AI Overviews pick their sources. This post is the first time ICX has put a number on its own position.
How was the September 2026 index measured?
Two engines answered every prompt on September 12, 2026.
The first was Perplexity Sonar, which runs a web search for every answer and returns sources. The second was OpenAI’s GPT-4o-mini with the OpenRouter web search plugin turned on, so it also searched the live web before answering. Both were called through the OpenRouter API with the same wording and a 700-token cap on the answer. Each prompt ended with one extra sentence asking the engine to include the URLs it used.
Every answer was scored two ways. Named means the text mentions Intelligent CX Consulting, ICX, or Christi Akinwumi. Cited means a URL on intelligentcxconsulting.com or christi.io appears in the answer’s sources. A short script does the scoring, so there is no judgment call. The whole run cost $0.37 in API credit and took about 20 minutes.
The limits are real and worth stating. This edition tested two engines, not five. Each prompt was run once per engine, so a single answer can swing a small count. Answers were collected from a US account with no personalization. ChatGPT’s consumer search, Google AI Overviews, Gemini, and Claude were not included. Later editions will add engines only when a full 28-prompt run can be repeated on them.
Who do AI engines name for conversational AI consulting?
The short answer: big firms on Perplexity, individual people on web-grounded GPT, and ICX in 5 of the 44 open answers.
Here is the scoreboard for the 22 category prompts, from the published data.
| Engine | ICX named | ICX cited |
|---|---|---|
| Perplexity Sonar | 1 of 22 | 0 of 22 |
| GPT-4o-mini, web search on | 4 of 22 | 3 of 22 |
| Combined | 5 of 44 (11%) | 3 of 44 (7%) |
On the six branded prompts, both engines named ICX and cited its pages in every answer. That is 12 of 12. Once an engine knows the name, it describes the firm accurately and links the about, services, and FAQ pages.
For the lead prompt, “Who are the best conversational AI consultants for enterprise customer service?”, Perplexity answered with Deloitte, Accenture, Kore.ai, Cognigy, Rasa, Zendesk, Salesforce Agentforce, and Parloa. Web-grounded GPT answered the same prompt with a list of named individuals, starting with a conversation designer whose personal site appeared in the sources of six of the 22 GPT answers. Both answers are in the published JSON.
The domains cited most often across all 44 category answers tell the same story. LinkedIn led with 17 source links spread over 8 answers. A freelance marketplace came next with 8 links in 2 answers. Then ICX’s own site with 7, a personal consultant site with 6, and a cluster of chatbot vendors at 4 to 6 each. The counts are source links, so one answer can list a domain more than once.
Where did ICX show up? On the two Dallas-Fort Worth prompts, GPT named the firm on both and cited the site on one. Perplexity named it on one, drawn from ICX’s LinkedIn company page rather than the ICX site. On the topical prompts, GPT cited the ICX post on enterprise chatbot best practices for the best-practices prompt, listing it five times in that one answer, and the post on intent architecture once. Perplexity did not cite an ICX page on any open question.
Why do individuals beat firms on these prompts?
This was the surprise in the data, and it matters more than ICX’s own score.
When a buyer asks “who are the top independent conversational AI consultants,” both engines answered with people, not companies. Perplexity named three individuals and said its list was cautious because the results were clearer on consultants and boutiques than on any established global ranking. GPT listed five people. Three of its five sources were LinkedIn profile pages and the other two were personal sites.
The pattern held across the hiring prompts. Engines that search the live web find LinkedIn profiles, personal sites, and marketplace listings, because those pages answer the question “who” directly. A firm’s services page answers “what,” so it is cited on branded prompts and skipped on discovery prompts.
ICX’s own results confirm it. The branded prompts about the founder returned christi.io pages and the ICX about page every time. The open prompts, where a name would have to be found rather than looked up, returned neither.
What does ICX’s own score say?
Plainly: ICX is invisible on Perplexity for open questions, partly visible on web-grounded GPT, and fully visible once someone already knows the name.
ICX was named once by Perplexity, on “Who provides conversational AI consulting in Dallas, Texas?”. That answer pulled the name from ICX’s LinkedIn company page, described the firm as an IT services company, and cited a competitor’s Dallas landing page instead of the ICX site. The three GPT citations came from two blog posts and the Dallas prompt. Nothing on the SaaS, ROI, or services clusters named ICX on either engine.
That is a 5 of 44 baseline. It is the number every later edition gets measured against.
What should a consultancy do with this?
Four things follow from the data, in order of weight.
First, the person is the citable entity. On “who” questions, engines cite LinkedIn and personal sites. A consultancy that hides its people behind a firm name is leaving the discovery prompts to whoever shows up as a person. The founder’s profile headline and about page should carry the exact words buyers type, and the firm’s site should name its author on every post.
Second, definitions and best-practice guides are the pages that get cited. ICX’s two open-prompt post citations both came from posts that define a term or list a checklist. The 80 other posts on the site earned nothing on these prompts. A page that defines a term or lays out a step-by-step checklist is more likely to earn a citation than a short opinion post, so the next posts in this series are built that way.
Third, fix the entity record. Perplexity described ICX as an IT services company because that is what the LinkedIn company page category said. An engine that finds a wrong label repeats it. The company page category and description get corrected before the next edition.
Fourth, measure it, then change one thing. This index costs under a dollar and 20 minutes per edition. A firm that runs it before and after a change knows whether the change worked. A firm that does not run it is guessing. The next edition will test whether adding an author entity and a Dallas-specific page moves the local and hiring prompts.
If you want ICX to run this scoreboard for your own brand and prompt set, that is a short project. See ICX services or the resources page for the tracking template.
How ICX verified this post
Every number in this post comes from one run collected on September 12, 2026. The raw answers, the URLs each engine returned, and the per-prompt scores are published as JSON and as a readable scoreboard. The scoring rule is a fixed text match on the brand names and domains, applied by a script, with no manual edits to the results.
Two flags. Each prompt ran once per engine, so a single answer can move a count by one. The engine list is two, not five, and ChatGPT’s consumer search was not among them. Both limits are stated so the next edition can be compared honestly.
ICX drafts with AI assistance and reviews every post before it goes live. Read the full policy on the AI disclaimer page. The index is scheduled weekly through October 19, 2026, then monthly, and each edition links back to this one.
Sources
Frequently asked questions
What is the ICX AI Citation Index?
It is a recurring public scoreboard. ICX runs a fixed set of 28 buyer questions through live AI engines and records whether ICX, or anyone else, gets named or cited. The prompt set, the raw answers, and the scoring script are published with every edition so anyone can check the numbers.
How often will the index be updated?
Weekly through October 19, 2026, then monthly. Each edition uses the same 28 prompts and the same scoring rule, so the trend is comparable across editions. New engines get added only when a full run can be repeated on them.
Why does ICX publish a score where it mostly loses?
Because the number is the point. Most AI visibility advice is opinion. This is a measurement, with the method stated, that a buyer or a competitor can rerun. A failing score published honestly is worth more than a winning score nobody can verify.
Which AI engines were tested?
Perplexity Sonar, which searches the web for every answer, and OpenAI's GPT-4o-mini with the OpenRouter web search plugin turned on. Both were called through the OpenRouter API with the same prompts and a 700-token answer cap. ChatGPT's own consumer search and Google AI Overviews were not tested in this edition.
Why do AI engines name individual consultants more than firms?
Engines that search the live web answer a 'who' question with pages that name a person: LinkedIn profiles, personal sites, and marketplace listings. A firm's services page answers 'what', so it gets cited on branded questions and skipped on discovery questions. In this run, LinkedIn was the most cited domain and the most named consultant was an individual with a personal site.
What does it mean to be cited versus named?
Named means the brand or person appears in the answer text. Cited means a URL on the brand's own site appears in the answer's sources. Cited is the stronger signal, because it means the engine read the brand's page and used it.
Get an email when we publish a new post.