Business news

AI Search and Software Stocks: How Often ChatGPT Recommends ADP, PayPal and HubSpot

Posted on October 2, 2026

By Bike Hawley

ChatGPT and other AI search tools now tell buyers which software to choose, and that makes their answers matter for software stocks such as Automatic Data Processing (ADP), PayPal (PYPL) and HubSpot (HUBS). The market has already shown it cares. On October 1, 2025, Reddit (RDDT) shares fell about 12% in a single session after data from tracking firm Promptwatch showed ChatGPT citing Reddit far less often than it had weeks earlier, Yahoo Finance reported. Advertising accounted for 93% of Reddit’s revenue in the quarter it had most recently reported. The stock moved anyway.

That was one company and one chart. ChatGPT reached 900 million weekly active users in February 2026, by OpenAI’s count, and when one of them asks which payroll or payments software to use, the answer names brands. To see which ones, High Salience, the AI search and SEO agency I founded, put 2,100 software-buying questions to ChatGPT in September 2026 and recorded every brand it named and every source it cited.

How the ChatGPT Answers Were Collected

The ChatGPT answers were collected between September 8 and September 26, 2026, across 21 software categories, with 100 buyer questions in each. The questions follow four patterns: “best CRM software,” “Salesforce vs HubSpot,” “Salesforce alternatives” and “which CRM should I use for a 5 person sales team.” Each was asked once, in the United States, in English, with web search turned on.

Two independent AI readers then coded every brand in every answer at two levels. A brand was recommended when the answer matched it to a stated need. It was picked when the answer made the brand its own choice, and one answer can pick several brands for different kinds of buyers. The readers agreed on 97.8% of their codes, and a third reader settled the rest.

Google’s top 10 organic results for the same questions served as a control. High Salience publishes the data for every category, so each count below can be checked.

Which Publicly Traded Software Companies ChatGPT Picks Most Often

Several publicly traded software companies are ChatGPT’s most frequent pick in their category. HubSpot was picked in 69 of the 100 CRM answers, compared with 37 for Salesforce (CRM). Intuit’s (INTU) QuickBooks was picked in 65 of 100 accounting answers. Shopify (SHOP) was picked in 69 of 100 ecommerce platform answers, Zoom (ZM) in 63 of 100 video conferencing answers and DocuSign (DOCU) in 62 of 100 e-signature answers. In project management, Asana (ASAN) was picked in 70 answers and monday.com (MNDY) in 54.

One brand per category tends to become the default. Across the 21 categories, the most-picked brand was picked in a median of 61 of 100 answers, with a range of 44 to 78. That position is a form of distribution. The average answer recommended about four brands, so a buyer who asks an assistant instead of scrolling a results page may never see the vendor in tenth place.

Where Private Companies Lead ADP, Paychex and PayPal

Private companies lead ADP and Paychex in payroll and lead PayPal in payment processing, at least in ChatGPT’s answers. Gusto, which is privately held, was picked in 61 of the 100 payroll software answers. ADP was picked in 23 and Paychex (PAYX) in 8. The gap holds when the question does not name the brand. On the 80 questions that never mentioned Gusto, it was still picked 49 times. ADP was picked 20 times on the 88 that never mentioned it.

Payment processing looks similar. Stripe, the private company that Bloomberg reported had dropped its pursuit of PayPal alongside buyout firm Advent in late August, was picked in 54 of the 100 payment processing answers. Block’s (XYZ) Square was picked in 34 and PayPal in 22.

Some caution is in order. The payroll questions lean toward small employers, and ADP’s picks included the questions about enterprise payroll, large companies and employees in several states. The numbers describe the small-business conversation better than they describe ADP’s enterprise base.

The citations offer a clue to how a default forms. ChatGPT cited gusto.com in 65 of the 100 payroll answers, and 48 of those 65 were answers to questions that never named Gusto, such as “ADP alternatives.” It cited adp.com in 20 answers and paychex.com in 10. Gusto publishes comparison pages about its rivals, and the model reads them.

Most AI Search Picks Go to Brands That Do Not Rank on Google

Most AI search picks went to brands that did not rank on Google for the same question. Of the 6,475 picks in the sample, 4,944, or 76%, went to brands whose own website was not in Google’s top 10 organic results for that question. At the broader recommended level, it was 6,270 of 7,997, or 78%.

The sources behind the answers show the same split. ChatGPT’s answers carried 6,443 citations, counted once per website per answer. Only 1,490 of them (23%) went to a domain that ranked in Google’s top 10 for that question, while 3,604 (56%) went to the software vendors’ own websites. None went to Reddit, which is in line with Promptwatch’s report that Reddit’s share of ChatGPT citations fell to about 0.5% in mid-August 2026 from 3.8% a few weeks earlier.

For anyone who follows software stocks, the practical point is that the Google ranking data many analysts already track will miss most of this. Who is winning in AI answers has to be measured in the answers themselves.

How Investors Can Use AI Search Data

Investors can use AI search data as a signal about distribution, not as a sales figure. A pick in a ChatGPT answer is not a purchase, and High Salience did not test whether any of these counts relate to revenue or share prices. What the data shows is which brands a buyer is told to consider. For a company that depends on being found, that is worth knowing.

A few checks cost nothing. Ask an assistant the questions a customer would ask, and ask more than once, because answers change between runs. Look at the sources it cites. If a rival’s website is doing the explaining, the rival is writing the brief. Then listen on the next earnings call for whether management measures the demand that arrives through AI assistants at all.

Limits of This ChatGPT Data

The limits of this ChatGPT data are worth stating plainly. Each question was asked once, and answers vary from run to run, so a brand missing from an answer was not observed in that run and may appear in others. The sample covers one country and less than a month. The coding was done by AI readers working from a written rulebook, not by a human panel, and only brands on lists fixed before collection were counted. Nothing here is a recommendation to buy or sell any security.

Even with those limits, the pattern is worth watching. In October 2025, a chart of ChatGPT citations coincided with a double-digit drop in Reddit’s stock. The same kind of data now exists for 21 categories of business software, and it shows private companies such as Gusto and Stripe as ChatGPT’s most frequent pick in payroll and payments, ahead of their publicly traded rivals. Whether that shows up in reported results is a question for the coming quarters. It is one worth asking on the next call.

Bike Hawley is the founder of High Salience. On the date of publication, Bike Hawley held no positions, directly or indirectly, in any of the securities mentioned in this article.

Sources

High Salience, “ChatGPT Software Recommendations: 21 Categories, 2,100 Answers,” 2026. highsalience.com/research
Yahoo Finance, “Reddit stock falls for second day as references to its content in ChatGPT responses plummet,” October 1, 2025.
TechCrunch, “ChatGPT reaches 900M weekly active users,” February 27, 2026.
Bloomberg, “PayPal Deal Talks End as Advent, Stripe Group Abandons Acquisition Effort,” August 28, 2026.
Promptwatch, “Reddit Citations Are Dropping in ChatGPT,” August 18, 2026.

About High Salience

High Salience engineers commercial visibility across Google, AI search and the third-party sources shaping buyer decisions. It publishes original research on how ChatGPT and Google pick brands, including 21 software teardowns built from 2,100 ChatGPT answers. High Salience sells services in the field this research covers.