By Gregg Kell, Founder and Editor, Spotlight on Startups | November 2026
Here is how to get recommended by ChatGPT as a B2B company. Do two things. Make the company website easy for AI to read, with plain facts on it. Then get other people to write about the company on sites the company does not control. Ads will not do the job.
The research is consistent on that last point. Muck Rack’s October 2026 report found that paid content earned 0.3 percent of AI citations. Earned media earned about 84 percent. Spotlight on Startups, the Earned Media Platform for B2B Founders, built the seven-step playbook below from that research.
The playbook needs no software and no agency. It works for ChatGPT and for Perplexity, although the two behave differently. Every figure here is current as of November 2026.
How to get recommended by ChatGPT: the short answer in four numbers
- 14.4 percent. The share of ChatGPT answers that involved a web search in an Ahrefs study of 66.6 million responses. ChatGPT answered the rest from what it already knew.
- 0.66 to 0.71. The correlation between brand mentions across the web and visibility in ChatGPT, Google AI Mode and AI Overviews, from an Ahrefs study of 75,000 brands.
- 84 percent against 0.3 percent. Earned media’s share of AI citations compared with paid content’s share, from Muck Rack.
- Less than 1 in 100. The chance that ChatGPT returns the same list of brands twice, according to SparkToro.
Those numbers set the strategy. ChatGPT mostly answers from memory, and its memory comes from what the web has said over time. So, the work is to be written about, and to keep being written about.
How ChatGPT and Perplexity build a shortlist
A buyer types a question such as which firms handle payroll compliance for mid-sized manufacturers. ChatGPT and Perplexity then build an answer in a few steps. Knowing those steps shows a founder where to act.
ChatGPT often answers from memory
Ahrefs studied 66.6 million ChatGPT responses from August and September 2026. Only 14.4 percent involved a web search. Ahrefs concluded that for many questions, a brand’s chance of appearing depends on what ChatGPT already knows about it.
That study covered free, logged-out sessions, so it is a snapshot. Buying questions trigger more searching. Grow and Convert tested 100 buying questions and found that 83 triggered a web search.
ChatGPT rewrites the question before it searches
When ChatGPT does search, it does not search the buyer’s exact words. Peec AI analyzed 5 million of these hidden searches in April 2026. ChatGPT added words such as best, reviews and the current year, even when the buyer never typed them. Peec AI sells tracking software, so treat the study as vendor research.
ChatGPT checks the company’s own website for facts
The Ahrefs study also looked at searches aimed at one specific website. About 76 percent of those went to a brand’s own site. ChatGPT uses outside sources to decide whom to name. It then uses the company website to confirm details such as pricing and features.
Perplexity searches every time
Perplexity works more like a search engine. Profound compared the hidden searches of three AI engines and found that Perplexity stays closest to the buyer’s original wording. Profound also found that 91 percent of ChatGPT’s searches were unique from one run to the next. Profound is a tracking vendor as well.
The practical difference is simple. A new article can show up in Perplexity quickly, since Perplexity looks at the live web each time. ChatGPT changes more slowly, since much of its answer comes from memory.
Why watching a dashboard changes nothing
More than forty AI visibility tools were on the market as of September 2026. They run a list of questions through AI assistants and report whether a company was named. Our review of what AI visibility tools actually do reached a blunt conclusion. The tools measure. None of them changes the answer.
A dashboard is useful after the work has started, in the same way a bathroom scale is useful after a diet has started. Buying the scale first does not help. A founder can get the same starting data by hand in an afternoon.
There is a second problem with early tracking. SparkToro ran 2,961 tests and found that the same brand list came back less than once in 100 tries. A tool that reports a ranking position is reporting noise. How often a company appears is the number that means something.
The seven-step playbook
The seven steps below run in order. Steps one to four fix what the company controls. Steps five and six build what it does not control. Step seven keeps score.
Step 1. Write down the buyer’s questions
List five to ten questions a buyer would ask before knowing the company exists. Leave the company name out. Use the buyer’s words, such as who is the best fractional CFO for a software startup. Add one or two comparison questions, since ChatGPT leans toward words like best and reviews.
Step 2. Take a baseline by hand
Ask each question in ChatGPT, Perplexity, Gemini and Claude. Run each one three times. Record how often the company appears and which competitors appear in its place. Open the cited sources and note which websites keep coming up.
Save the results in a spreadsheet. This is the starting line. Our guide to why ChatGPT recommends competitors explains how to read the gaps.
Step 3. Make sure AI can read the website
Vercel studied AI crawler traffic and found that none of the major AI crawlers run JavaScript. A page that loads its text in the browser can look empty to them. View the page source and search for a sentence from the main text.
Check for accidental blocking too. Playwire reported in June 2026 that a Cloudflare setting called Block AI bots was on by default. Review that setting and the robots.txt file.
Step 4. Put plain facts on the website
ChatGPT visits the company website to confirm facts, so give it facts. State in one sentence what the company does and who it serves. Name the industries, the locations and the specific services. Answer common buyer questions on the page in plain words.
Three pages do most of this work. An about page says who runs the company and when it started. A services page names each service and who it suits. A page of common questions answers what buyers ask on sales calls, in the buyer’s own words.
Vague claims do not help. A sentence like trusted partner for growing businesses gives ChatGPT nothing to repeat. A sentence like payroll compliance for manufacturers with 100 to 500 employees in Ohio does.
Step 5. Get written about by independent publications
This step carries the most weight. Ahrefs studied 75,000 brands across ChatGPT, Google AI Mode and AI Overviews. Mentions on other websites correlated with AI visibility at 0.66 to 0.71. Ahrefs notes that correlation does not prove cause.
Reported articles count most. BuzzStream analyzed 4 million AI citations and found that 81 percent of news citations went to original editorial content. A B2B founder has four realistic routes:
- National business press such as Forbes and TechCrunch, for companies with funding or major news.
- Trade publications that the company’s buyers already read.
- Journalist request platforms such as Qwoted, where a founder can earn a quote in someone else’s story.
- Founder publications such as IdeaMensch and Spotlight on Startups.
Step 6. Join the lists and the videos
Buyers ask for the best options, and ChatGPT searches for best-of content. BuzzStream found that list and review articles made up 39.07 percent of news citations for evaluative questions. Find the real comparison lists in the industry and ask the authors to consider the company.
Video matters more than most founders expect. The same Ahrefs study found that mentions on YouTube had the strongest correlation of any factor, at about 0.737. A podcast interview that also posts to YouTube counts. So does a recorded conference talk.
Step 7. Retest every month and keep content fresh
Repeat the baseline from step two each month. Look at how often the company appears, and ignore small swings. Refresh the key pages on the website with current dates and figures.
Recent content has an edge. Muck Rack’s December 2025 report found that half of all AI citations pointed to content published in the previous 11 months. One article from three years ago will not carry a company for long.
The playbook at a glance
The table below lists the seven steps with the action and the evidence for each.
What does not get a company recommended
- Advertising. Ahrefs measured a 0.215 correlation between branded ad spend and AI mentions. Muck Rack put paid content at 0.3 percent of citations.
- Publishing more pages. Ahrefs found that the number of pages on a site had one of the weakest links to AI visibility, at 0.17.
- Wire press releases. BuzzStream put syndicated releases at 0.04 percent of AI citations. See our guide on whether press releases are worth it for AI search.
- Tracking alone. A dashboard reports the problem. It does not fix it.
Five mistakes that waste the first month
Founders who run this playbook tend to lose time in the same five places. Each one is easy to avoid.
- Testing with the company name in the question. A buyer who already knows the name is not the buyer the company is missing. Our benchmark found that all 30 companies tested were described correctly once named.
- Testing once. A single answer is one roll of the dice. Run each question at least three times before drawing a conclusion.
- Testing while logged in. An assistant may tailor answers to the account. Use a logged-out or private browser window so the results match what a stranger would see.
- Counting one mention as a win. SparkToro’s data shows that lists change constantly. Look for a rising share of appearances across a month.
- Stopping after one article. AI assistants look for agreement across several independent sources. One feature starts the record. It does not finish it.
How long the playbook takes
No one can promise a date. Steps one to four take a founder a few days. Steps five and six depend on other people, so they take longer. Perplexity can pick up a new article soon after it is published, since it searches the live web.
ChatGPT moves more slowly. With only 14.4 percent of answers involving a search in the Ahrefs data, much depends on what ChatGPT already knows. Plan in months and expect steady gains in how often the company appears.
How Spotlight on Startups fits into the playbook
A disclosure belongs here, since this section describes a product I run. Spotlight on Startups is the Earned Media Platform for B2B Founders, a journalist-authored publication that publishes reported founder features structured so ChatGPT, Perplexity, Claude, and Google AI Overviews can find, understand, and cite them. I launched Spotlight on Startups in 2025.
A founder feature is one way to complete step five. Spotlight on Startups charges nothing for it. The process is a 10 minute Zoom call and five written questions. A writer then reports and writes the story, and the publication controls the final copy. Our editorial standards explain how we select and check stories.
One feature is one source, and one source is not enough. Our 2026 B2B AI Citation Benchmark tested 30 small and mid-sized B2B companies. Not one was named when a buyer asked a general question. All 30 were described accurately once the buyer gave the company name. Their websites worked. Their outside proof was missing.
Founders can read how the free founder spotlight works or visit the Get Featured page. For background, see why ChatGPT may not know a startup exists, our overview of earned, paid and owned media, and the six question AI Ask Panel.
Where Kell Solutions fits
A second disclosure. I am also the CEO of Kell Web Solutions, which sells answer engine optimization services as Kell Solutions. A founder feature is never conditioned on buying anything.
Kell Solutions supports steps three, four and seven. Kell Solutions writes the content and delivers the recommendations. The company’s own team makes the changes to its website. The services are AI Crawler Access Setup, Signal AEO, Signal Authority and AEO Elite. Details and pricing are on KellSolutions.com, not on this publication.
Frequently asked questions
How do I get recommended by ChatGPT?
Get written about on websites the company does not control and keep the company website readable and factual. Ahrefs found that web mentions correlated with AI visibility at 0.66 to 0.71 across 75,000 brands. Reported articles, trade coverage, comparison lists and video interviews all add to that record.
Do ads help a company get recommended by ChatGPT?
Ads do very little. Muck Rack found that paid content earned 0.3 percent of AI citations, while earned media earned about 84 percent. Ahrefs measured a weak 0.215 correlation between branded ad spend and AI mentions. Ad budgets are better spent earning independent coverage.
How does Perplexity choose which companies to recommend?
Perplexity searches the live web for every question and builds its answer from the pages it finds. Profound found that Perplexity stays closest to the buyer’s original wording. New articles can appear in Perplexity answers quickly, so recent independent coverage matters a great deal.
Does ChatGPT search the web every time?
No. Ahrefs studied 66.6 million ChatGPT responses and found that only 14.4 percent involved a web search. ChatGPT answered the rest from memory. Buying questions trigger more searches. Grow and Convert found that 83 of 100 buying questions led ChatGPT to search the web.
Do I need an AI visibility tool to get recommended?
No. AI visibility tools measure whether a company is named. They do not change the answer. A founder can collect the same starting data by hand in an afternoon. A tool becomes worth paying for once the checks turn into a monthly routine.
How long does it take to get recommended by ChatGPT?
There is no fixed timeline. Perplexity can surface a new article soon after publication. ChatGPT changes more slowly, since most of its answers come from memory. Plan in months and track how often the company appears. Fresh content helps. Muck Rack found that half of citations are under 11 months old.
What is the first step for a small B2B company?
Write down five questions a buyer would ask, then ask them in ChatGPT, Perplexity, Gemini and Claude three times each. Record who gets named. That baseline costs nothing, takes about 30 minutes, and shows which competitors and which websites the company needs to match.
Sources and retrieval dates
All sources below were retrieved on 14 November 2026, except the two Muck Rack reports, the BuzzStream study and the Ahrefs May 2025 study, which were retrieved on 13 November 2026.
- Ahrefs, How ChatGPT Does Query Fan-Out (15 Million Queries Studied), 12 October 2026. Free, logged-out sessions only.
- Ahrefs, Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews, 12 December 2025.
- Ahrefs, An Analysis of AI Overview Brand Visibility Factors, 26 May 2025.
- PPC Land report on Peec AI’s study of 5 million fan-out queries, April 2026 data. Vendor research.
- Grow and Convert, fan-out query study of 100 buying prompts.
- Profound, What AI Engines Actually Search For. Vendor research.
- SparkToro and Gumshoe, AI brand recommendation consistency, January 2026.
- Unite.AI report on Muck Rack’s October 2026 edition and Muck Rack, What Is AI Reading?, December 2025.
- BuzzStream, The Role of News Publications in AI Citations, 9 March 2026.
- Vercel, The Rise of the AI Crawler and Playwire on Cloudflare’s AI bot setting, 15 June 2026. Cloudflare’s defaults have changed more than once.
Corrections are welcome at [email protected]. Spotlight on Startups can also be reached at (949) 354-6605.