HIGH SALIENCE / RESEARCH
AI Search Research: How ChatGPT and Google Pick Brands
High Salience's AI search research is a set of original studies of how ChatGPT and Google pick brands: the way search engines and AI systems select, cite and recommend them. We don't repeat search theory. We test it. Every study follows the same discipline: claim, evidence, methodology, data, limitations, conclusion.
How ChatGPT recommends 187 software brands
One data page per brand that appeared in 10 or more teardown answers: how often ChatGPT names it, recommends it for a stated case and makes it its own pick, who ChatGPT says it is for, head-to-head results, what it suggests instead, and the sources behind each recommendation.
AI answers in four industries: ChatGPT and Google read different lists
100 buyer questions in fintech, healthcare, legal and ecommerce, asked to ChatGPT and Google on the same day. Most of the sites ChatGPT cited did not rank in Google's top ten, and ChatGPT cited Reddit and YouTube in none of its answers.
100 AI writing tool questions: the model's own product is picked most, and specialists win the named uses
ChatGPT is picked in 50 answers, ahead of Jasper (35) and Grammarly (34), a result to read knowing ChatGPT is the one answering. Sudowrite, Shopify Magic and Surfer win their named uses, and the model cites 143 different sites, the most in the series.
100 SEO tool questions: two suites share the verdicts, specialists keep their lanes, and Reddit is never cited
Ahrefs is picked in 55 answers and Semrush in 46. Reddit ranks in Google's top ten for 90 of the questions and gets no citations; ahrefs.com ranks for 5 and is cited in 42 answers.
100 marketing automation questions: HubSpot leads a second category, and its own website is cited in half the answers
HubSpot is picked in 57 answers, ActiveCampaign in 45. hubspot.com is cited in 50 answers, 37 of them to questions that never mention HubSpot. Industry questions add a specialist: Klaviyo, HighLevel, Lawmatics, Kit.
100 e-signature questions: DocuSign is the default pick, Adobe is the pick for PDF work, and Dropbox Sign is the way out
DocuSign is named in 67 answers and picked in 62, including 19 of 20 advice answers. Dropbox Sign is picked in 7 of 10 answers about replacing DocuSign. Google ranks adobe.com for one question; ChatGPT cites it in 35 answers.
100 video conferencing questions: Zoom is the default, and Teams and Meet are picked for their suites
Zoom is named in 76 answers and picked in 63. Microsoft Teams is picked in 41 and Google Meet in 35, mostly for buyers already on Microsoft 365 or Google Workspace. Webinar questions keep Zoom and add a specialist.
100 scheduling software questions: Calendly is the default, and the trades get their own picks
Calendly is named in 57 answers and picked in 51; no other brand is named in more than 35. In 11 of 20 trade questions none of the three most-picked brands is picked, and three roundup sites that rank for none of the questions are cited in 17 answers.
100 website builder questions: Wix is named most, and Squarespace wins the industry questions
Wix, Squarespace and WordPress lead, picked in 62, 56 and 47 answers. Squarespace is the pick in 15 of 18 industry questions. The sites Google ranks most for these questions, Reddit included, are cited in none of them.
100 ecommerce platform questions: Shopify is picked in 69 answers, and its own site is the most-cited domain in the series
Shopify appears in 75 answers and is picked in 69, 29 more than WooCommerce. shopify.com is cited in 67 answers, more than any domain in any teardown so far, 47 of them to questions that do not mention Shopify.
100 data warehouse and ETL questions: three warehouses picked together, and Snowflake as everyone's example destination
Snowflake is picked in 46 answers, BigQuery 43, Databricks 37. In 32 answers about pipeline tools Snowflake is only the place the data goes. snowflake.com ranks in Google's top ten for 2 of the 100 questions, yet it is the most-cited domain. Vendors' own sites supply 74 percent of citations.
100 business intelligence questions: Power BI is the default pick, and the vendor Google ranks most is chosen only when named
Power BI is named in 75 answers and picked in 69. Domo's site ranks in Google's top ten for 43 of the questions; Domo is picked in 3, all of them questions that name it. 87 percent of picks go to brands that do not rank for the question, the highest in the series so far.
100 endpoint security questions: Microsoft Defender is picked as the product buyers already own, and picks come as shortlists to test
Microsoft Defender appears in 68 answers and is picked in 62, nearly always tied to Microsoft 365 licensing. CrowdStrike, SentinelOne and Sophos follow. Two independent test labs are cited in 11 answers without ranking for any of them.
100 password manager questions: the strongest default in the series, and LastPass has almost disappeared
Bitwarden is named in 84 answers and picked in 78, more than any brand in any earlier teardown. Vendors' own sites supply 69 percent of citations. LastPass is named in 2 of the 89 answers to questions that do not mention it.
100 applicant tracking questions: four leaders, clean lanes, and the widest gap from Google yet
Workable, Greenhouse, Ashby and Lever lead. 86 percent of recommendations go to brands whose own site does not rank for the question, the highest in the series so far, and a comparison site that ranks for none of them is cited in 13 answers.
100 HR software questions: three brands share the top, and one comparison page is cited in 13 answers
Rippling, BambooHR and Gusto split the picks almost evenly. A single page comparing exactly those three is cited in 13 answers and ranks in Google's top ten for none of them. 17 answers make no pick at all.
Google's AI Overviews cite YouTube most, and ChatGPT never did
Same 700 questions. YouTube supplied 524 of 2,395 AI Overview source links and appeared in 221 of 647 Overviews. ChatGPT's answers cited it zero times, and took about half their citations from vendor sites against one in six for the Overviews.
Reddit ranks for 90% of software questions in Google, not in ChatGPT
Same 700 questions, same day. Reddit ranked in Google's top ten for 631 and the top three for 515. Google's AI Overviews cited it in 112 of 647. ChatGPT's live web-search answers mentioned Reddit zero times, though aggregated ChatGPT data tells a different story.
100 payroll questions: one brand in every recommendation pick, and its website in two answers out of three
Gusto is picked in all 20 recommendation queries. gusto.com is cited in 65 of 100 answers, 48 of them for questions that never mention Gusto. Paychex is named in 28 and picked in 8. Most answers ask for your headcount first, and 19 make no pick at all.
100 payment processing questions: one default, then a rate card
Stripe appears in 70 of 100 answers, is recommended in 68 and picked in 54. Eighty-four answers ask for the buyer's own numbers, 63 quote a published rate, and 22 pick no processor at all. A category outside the SaaS benchmark, under the same codebook.
100 accounting questions: the incumbent survives every vertical prompt
QuickBooks appears in 75 answers, is recommended in 73 and picked in 65. Law firms get Clio plus QuickBooks, restaurants get MarginEdge plus QuickBooks, stores get A2X plus QuickBooks. Forty-four percent of recommendations go to brands that rank, the most of any category so far. Five categories side by side.
100 help desk questions: Zendesk leads, and a site nobody knows is the third most-cited source
Zendesk appears in 70 answers, is recommended in 69 and picked in 61. Getmacha.com is cited 20 times, ahead of Freshworks and G2. The vertical prompts did not swap the field. Four categories side by side, recoded under codebook v2.0.
100 email marketing questions: seven brands and no default
No brand appears in more than 58 of 100 answers. Every one of 30 head-to-heads recommended both sides. Vertical prompts add a specialist beside the general brands instead of replacing them. Three categories now side by side under one protocol.
100 CRM questions, same method: ChatGPT picks from a big four
HubSpot is named in 75 answers and picked in 69; Pipedrive, Zoho CRM and Salesforce are picked in most of the answers that name them. Every Salesforce recommendation comes with a big-company condition. Real estate and law prompts cleared the general field for vertical tools.
100 questions, one category: what ChatGPT actually recommends
Asana is named in 73 of 100 answers and picked in 70. In project management nearly every brand ChatGPT names comes with a stated case, all 30 head-to-heads are forks, and 85 percent of recommendations go to brands whose own site does not rank in Google's top ten.
AI SEO Agencies Compared: The 2026 Landscape
The agencies selling GEO, AEO and AI SEO, compared on what they actually publish: specialization, disclosed pricing, proprietary research, documented results and methodology transparency. High Salience is included and the publisher is disclosed.
1,000 commercial SaaS queries across ChatGPT, Gemini and Google AI Mode
Which brands get recommended, which sources get cited and how much the three surfaces disagree. Methodology published, query set approved and frozen. Findings appear when the benchmark has run, not before.
One dataset, several studies. The two below are cut from the Benchmark 01 data rather than collected separately, and publish after it.
The 100 domains influencing enterprise software recommendations most frequently
A citation census across AI surfaces: which third-party domains keep shaping the answer, and how concentrated the influence really is.
Does ranking in Google predict citation in AI answers?
The study tests how strongly classic organic position correlates with AI citation across the controlled query set.
Why we publish
Why High Salience publishes AI search research
High Salience publishes AI search research because research is the moat. It sits in the middle of a flywheel. Client work produces evidence. Evidence produces studies. Studies produce citations, links and the trust of the people and systems that shape categories. That flywheel is how High Salience earns the authority we build for clients, and it's why our AI search optimization and AEO strategy work runs on tested findings instead of borrowed theory.
The findings feed straight into client work: every AI SEO engagement we run is measured with the same approach the teardowns use.
New to the topic? Our AI search guides explain the basics, starting with how AI search engine optimization works.
Journalists and analysts: every study includes methodology and, where possible, a downloadable dataset. If you need something clarified before citing, ask.
The Edge
What is decision latency in search?
Decision latency is the time it takes to turn a change in search into the correct commercial decision. Search changes continuously. The firms relying on public consensus learn what happened after everyone else has already read the same analysis.
High Salience is built to shorten that gap. We monitor markets, preserve historical evidence, compare winners against losers, test hypotheses and feed what survives back into strategy.
The objective is not more SEO information. It's getting to the correct commercial decision earlier.