Which integrations does AI name, and do they become recommendations?

An answer might recommend a product because it “integrates with Salesforce.” That names Salesforce as a partner, but does not necessarily recommend it. We counted each partner once per answer and checked whether the answer also presented it as a separate option.

In this sample, we found 976,159 partner mentions in 481,577 answers. 636,072 of them, 65.2%, were integration-only. The share was 63.1% to 69.6% on every engine. Most of the time, a brand named as an integration is context for another brand's recommendation, not a recommendation itself.

Across 5,367,238 answers to organic prompts, 481,577 (9.0%) named at least one integration partner. ChatGPT did so most often, in 10.1% of answers. ChatGPT Search did so least, in 6.7%.

Answers that named partners usually named more than one. The average was 2.0 partners. 43.7% named exactly one and 27.1% named three or more.

The question decides most of it. Answers to prompts with integration wording, such as "integrates with Shopify POS", named a partner 45.1% of the time. Answers to all other prompts did so 7.9% of the time.

Salesforce was named as an integration partner in 46,173 answers, Slack in 44,400 and QuickBooks in 38,165. AWS and GitHub followed, each above 36,000.

The list is concentrated. The 10 most-named partners account for 35.2% of all partner mentions and the top 30 for 61.8%. Collaboration tools made up 29.0% of partner mentions and developer and cloud platforms 26.6%.

The last column shows how often each partner was integration-only. Microsoft 365 was named as a partner in 23,659 answers and was integration-only in 89.3% of them. QuickBooks was integration-only in 50.9%, because AI also recommends QuickBooks itself as accounting software.

Among the 146 partners named in at least 1,000 answers, some were integration-only nearly every time. Siri was integration-only in 94.7% of the answers that named it as a partner, MySQL in 92.7% and Active Directory in 90.5%.

The large office suites follow the same pattern. Microsoft 365 was integration-only in 89.3% of 23,659 answers, Gmail in 88.7% of 10,194 and Outlook in 85.1% of 14,981. AI treats these brands as the systems a buyer already runs, and recommends something that connects to them.

Some brands are named as integrations and recommended in the same answer. RingCentral was integration-only in 13.3% of its 1,072 partner answers, Rippling in 24.0% and Paychex in 28.5%. ClickUp, FreshBooks, Zoho and Toast were integration-only less than a third of the time.

These products sit on both sides of a connection. AI recommends them in their own category and also names them as the system another tool connects to. For these brands, being named as a partner and being named as an option go together.

For every partner mention we found the brand the integration belongs to: the brand in the same sentence, or the brand whose list item the sentence sits in. We counted the prompts where each pair appeared and removed pairs from the same company, such as Microsoft products connecting to each other.

HubSpot was the brand most often named with a Salesforce integration, on 116 prompts. ClickUp, Rootly and Notion were the brands most often recommended with a Slack integration. Ramp, Relay and Brex led for QuickBooks, and Klaviyo for Shopify. Finance tools cluster around accounting platforms, and work-management tools around Slack, Jira and Microsoft Teams.

Integration breadth is the number of different brands an AI named with a given partner on at least two prompts. Salesforce had the widest set, with 1,213 brands. Slack had 1,020 and HubSpot 817.

A wide set means the platform is a common requirement across many markets. It also means each brand that connects to it shares the mention with hundreds of others. Saying that you integrate with Salesforce is rarely what separates one product from another.

We grouped prompts into the markets of the public Parse index and kept the 1,425 markets with at least 1,000 answers and 3 organic prompts. In 45 of them, more than half of answers named a partner. In 774, fewer than 1% did.

CRM Telephony Integration Software led at 86.1%, with Salesforce and HubSpot named in most answers. Employee Recognition and Rewards Platforms followed at 85.2%, almost always with Slack. Integration partners matter most in software that lives inside another tool, and barely at all in consumer goods.

ChatGPT named a partner in 9.4% to 11.6% of answers each month from November 2025 to March 2026. Google AI Overviews stayed between 6.3% and 9.3%.

On the current engines, ChatGPT Search fell from 8.6% in June 2026 to 5.0% in August and 5.1% in September. Google AI Mode stayed between 7.6% and 9.3%. The ChatGPT Search drop starts the month OpenAI moved ChatGPT to gpt-5-6, on August 8. The prompt mix also changes from month to month, so treat the timing as a lead, not a cause.

We followed 3,975 cases where a partner brand was first named on an organic prompt as integration-only, on a prompt we had already watched for 30 days and kept watching for 90. In 177 cases, 4.5%, a later answer to that prompt named the brand as an option. For the same brands on other prompts in the same market where they had not been named, the rate was 2.7%.

With a looser test, a later counted brand mention of any kind, the rates were 26.0% against 9.3%. Most of those later mentions were still integration context. Being named as a partner makes a later recommendation more likely, but it rarely leads to one.

We compared 436 prompts with integration wording to the other prompts in their market. The brand named most often on the integration prompt was the market's usual winner 28.9% of the time. For ordinary prompts compared the same way, it was 30.2%. Adding "integrates with Shopify" changes the winner, but no more than any other change of question does.

The platform named in the prompt was itself counted as a brand mention in 37.8% of 22,931 answers to those prompts. When a buyer asks for a tool that works with Slack, Slack is named in the answer, but usually as the platform, not the pick. For example, Bonusly was the brand named most for "employee recognition platforms that integrate with Slack", in 88.0% of answers.

An integration page is a page whose address contains /integrations or /connectors. An app marketplace listing is a page on a store such as apps.shopify.com, the Salesforce AppExchange or the HubSpot ecosystem. Among 117,304 answers with citations to organic prompts with integration wording, 22.2% cited at least one of these. ChatGPT did so in 35.1% and Google AI Mode in 13.1%.

The most-cited hosts were zapier.com, marketplace.atlassian.com and apps.shopify.com. Brands whose own integration page was cited on a prompt were named in 26.2% of that prompt's answers on average. Other brands named on the same prompts averaged 5.7%. The citation and the mention arrive together, so this is an association, not proof that the page caused the recommendation. It is a measured version of claims that integration pages raise recommendations by a fixed multiple, which we could not find a source for.

Overlap is the share of names two answers have in common, out of all names either one used. We paired answers to the same prompt on the same day where both engines named a partner.

ChatGPT and Google AI Overviews had 46.2% overlap in the partners they named, and 26.4% in the brands they mentioned. ChatGPT Search and Google AI Mode had 44.7% and 29.3%. About one pair in five named exactly the same partners. The same engine on consecutive days overlapped 54.7% to 64.2% on partners, so the engines disagree only a little more than one engine disagrees with itself. Engines agree on which platforms matter for a question more than they agree on what to buy.

We read the full text of every answer to an organic prompt from October 3, 2025 to October 6, 2026. Organic prompts are the questions Parse asks on its own; prompts that customers track for their own brands are excluded from every number. The engines are ChatGPT and Google AI Overviews to April 2026, ChatGPT Search and Google AI Mode from May 2026. In each sentence that follows an integration phrase ("integrates with", "integrations:", "connects to", "syncs with", "works with", "compatible with") we looked for 165 named platforms, from Salesforce to Siri. The list came from the most frequent names after those phrases and leaves out generic terms (CRM, API), protocols and operating systems, and bare company names such as Google or Microsoft. Brands outside the list are not counted, so the totals are a floor.

We hand-checked 120 random sentences: 197 of 203 detected partners, 97.0%, were real integration partners. The errors were options listed after an integration phrase and names picked up from link text. A second check used Parse's answer judge on 83,848 partner mentions from August 7 to October 5, 2026: of the mentions our rule treated as options, the judge rated 43.6% as not offered. That puts the share of partner mentions that are not recommendations closer to 80% in that window, so 65.2% is a conservative figure.

Brand mention tests exclude March 30 to April 25, 2026, when brand detection failed. Engine pairs are reported separately because the four engines cover different months.

Separate integration mentions from recommendations. Being named as a compatible tool is useful exposure, but it is different from being offered as the product to buy. Track both so you can see where your brand is becoming a choice.

If you sell an add-on, the platforms you connect to decide which questions you can appear in. Brands that were named as partners were 1.7 times as likely to become an option on that prompt later, and answers to integration questions cite integration pages and marketplace listings 22% of the time. Publish one page per integration with the platform's name in it, and keep your marketplace listings on Shopify, Salesforce, HubSpot and Zapier current.

If you are the platform, expect most mentions to be context. Microsoft 365, Gmail and Outlook were integration-only in more than 85% of their partner mentions. For these brands, a high mention count says little about how often AI recommends them.