Updated on 7 November 2026, by João Leite, Founder of murmur.
To appear in ChatGPT in software comparisons, the site needs to state, in text, what the AI compares: who it is for, price and integrations. When the buyer asks for the best system or compares two products, the AI searches the web and builds the list from what it read. The seller’s site is among the pages it reads at that moment. The page that answers the comparison, with the product facts in plain sight, is what the AI can read to build the list.
- The number of names per answer changes from one sector to another. In the legal software study, from September 2026, there were 4.8 names per answer, across 99 answers, the sector with the most names among those measured. In the marketing automation one, the AI cited only one Brazilian tool in 41 of 75 answers.
- Some of the answers do not cite any software from the sector list. Across nine software studies, they ranged from 10 of 75 answers in the real estate study to 38 of 75 in customer experience.
- In the software studies, the vendor’s name almost always appears when the AI read its site. In the BPM one, business process management software, 36 of the 75 answers, from ChatGPT and Claude, showed the sources. In those, the vendor’s name appeared in 97% of the times the AI read its site, and in 9% when it did not.
- In the real estate system study, the buyer profile brought more generic CRM. Generic CRM is the customer management system built for any sector. It appeared in 3 of the 15 answers to the general question (20%) and in 7 of the 21 that stated the profile (33%).
- In some of the “best software” questions, the Google AI summary does not even appear. That is what SEO consultant Lily Ray saw in the AI Overview, also called Google AIO: about one in five of the 100 questions she analyzed, in English, from April to June 2026, did not bring the summary.
murmur, a GEO agency in Brazil that places brands in the answers of ChatGPT, Claude, Google AIO and other AI agents, studies each client’s sector. It writes the content that answers the buyer’s questions and tracks the result week after week. In this guide, murmur does not rank any software. GEO is the work of making the company appear in the answers of ChatGPT and the other AI models. In September 2026, murmur produced and published sector studies of software, each with 25 buyer questions and no company names. The questions went to ChatGPT, Claude and Google AIO, and the legal software ones also to Perplexity. Selling to companies in general, from the buying committee to the shortlist of vendors, is covered in how to do GEO for a B2B company.
When the AI builds a software comparison, and when the answer comes out without any software from the sector
The AI builds a comparison when the question asks for a choice, and even then some of the answers come out without any software from the sector list. In murmur’s studies, each sector’s list gathers the Brazilian software that the AI cited at least once. In the bidding and marketing automation studies, the AI also cited procurement portals and foreign tools, which stay out of the list.
| murmur study, September 2026 | answers without any software from the list |
|---|---|
| real estate system | 10 of 75 |
| BPM, process management | 13 of 75 |
| payroll | 14 of 75 |
| time and attendance | 15 of 73 |
| contract management | 15 of 75 |
| marketing automation | 16 of 75 |
| bidding software | 18 of 74 |
| dialer and call center | 21 of 68 |
| customer experience platform | 38 of 75 |
In Google’s AI Overview, there are “best software” questions that do not even get a written answer from the AI. SEO consultant Lily Ray analyzed 100 questions of this kind, about software for companies, in English, between April and June 2026. The answers came from Brand Radar, an Ahrefs tool. About one in five did not bring an AI Overview.
How many software products enter a comparison answer, and in what position
In murmur’s legal software study, from September 2026, 4.8 software products entered per answer, on average, across 99 answers. The answers came from ChatGPT, Claude, Google AIO and Perplexity, and it was the fullest sector murmur measured. In the marketing automation study, the AI cited only one Brazilian tool in 41 of 75 answers. In legal software, the position of each name also matters: the system that most often opened the list was, on average, in 1.9th place. Another appeared in 13 answers, but came in at 6.8th place, after several names.
In the open-ended question, answered by someone who just wants to know which is the best, the most cited system in the legal software study appeared even more often. It was 96% of the 23 answers of that kind, against 70% of the 99 answers in the whole study. Each AI also returns its own list, as the list of companies changes from one AI to another shows.
In the BPM, legal software and real estate system studies, the vendor’s site is among the most read
In the BPM, legal software and real estate system studies, the AI read mostly the seller’s site. When a site was read, its owner almost always made it into the answer. In the bidding software study, the most-read site was gov.br.
In murmur’s BPM study, 36 of the 75 answers, 24 from ChatGPT and 12 from Claude, showed the sites consulted, and 25 appearances among the ten most read sites were from vendors. There, the vendor’s name was in the answer in 97% of the cases where the site was read and in 9% of those where it was not. In the legal software study, it was 84% against 17%, across 50 answers from ChatGPT, Claude, Google AIO and Perplexity with a visible source. In the real estate system study, it was 99% against 9%, across the 36 answers from ChatGPT and Claude that showed the sources.
In all three, reading the site and citing the name went together. What the AI reads on the vendor’s site, and what carries weight outside it, in reviews and in the press, is covered in where the AI checks the B2B vendor: its site, reviews and the press. Among the sources of the comparison, the vendor’s site is the only one the company controls, and whatever it does not say, the AI goes looking for on other sites. Which type of site carries more weight in ChatGPT’s sources, from the company site to the press, is the topic of sites that influence ChatGPT in Brazil.
Comparison page and alternatives page: what each one needs to have
The comparison page and the alternatives page need to state, in text, who each product serves and when it is worth switching one for the other. The comparison one answers “X or Y”, and the alternatives one answers “alternative to X”, with the words the buyer types. A page that only praises its own product does not answer either of the two questions.
A fact about a competitor comes from the competitor’s own page, with the date it was read. Another company’s price and plan change without notice, and a wrong statement about a competitor is a legal risk. One page per competitor, with the same text and only the name changed, adds nothing. Pages mass-produced to rank in search, rather than for the person reading them, are what Google calls scaled content abuse in its spam policy.
What the page needs to have to appear in ChatGPT in the comparison: who it is for, price and integrations
The page needs to include, in text, each point the buyer compares between software products: who it is for, price, plans, integrations and switching systems. In the legal software study, the questions talk about cost benefit, integration with the courts and a mobile app. To appear in ChatGPT in those questions, each point calls for a page or a section. The criterion is the software buyer’s question: one that actually changes calls for its own page, with the complete answer, and one that only changes wording stays on the same page.
- Who it is for: one page per profile served, with the profile in the title, such as “para imobiliária de locação” (for a rental real estate agency).
- Price and plans: the starting price and what each plan includes, or, where no price is published, what it depends on, such as users, modules and volume.
- Integrations: one page per integration customers look for, saying what passes from one system to the other.
- What is left out: the limit of each plan and what is paid separately.
- Switching systems: what migrating is like, how long it takes and what comes over from the old system.
Customer profile, price, plans and integrations are in text on the page, not in an image, a PDF or a tab that only opens on click, because for AI crawlers to read the site, the text must come ready in the HTML.
In the real estate system study, the question that says who uses it brought more generic CRM
When the question says who will use the software, such as “for an independent broker”, more generic CRM appeared in the answer. In murmur’s real estate system study, some real estate system appeared in every answer to the general recommendation and to the questions with a profile. Some generic CRM, not made for real estate, appeared in 20% of the 15 answers to the general recommendation. In the 21 answers to the questions that stated the profile, such as a rental, new-development or high-end real estate agency, it appeared in 33%.
In murmur’s software studies, the software list appeared less in the question from someone still learning how to buy. In the bidding software study, some software appeared in 87% of the 15 answers to the general recommendation and in 40% of the 15 that ask how to choose and how much it costs. In those 15, some purchasing portal, public or private, appeared in 73%. In eight of murmur’s software studies, with 15 answers of that kind each, or 14 in the dialer one, the AI cited 1 to 9 software products from the list, depending on the sector.
Each profile the product serves calls for a page with that buyer’s words, and that is where the site says who the product is for. The breakdown by company size and stage, starting with the startup competing for space with the leader, is in how to do GEO for startups.
The “best” list on your own site and the buying guide
The “best” list on your own site serves the buyer when it works as a guide to how to choose. That guide compares the options by what each one publishes and applies the same criteria to its own product. It is the format murmur uses in GEO agencies in Brazil: what each one publishes. The store that sells a product builds the buying guide by buyer profile, as what the AI reads before recommending a product shows.
In the real estate system and BPM studies, software comparison sites trailed vendor sites
In the real estate system and BPM studies, software comparison sites appeared among the sources, behind vendor sites. In 36 answers from ChatGPT and Claude with a visible source in each, the most-read vendor site appeared in 17 and in 9, and the most-read comparison site, in 6 and in 4. Google AIO does not show the list of sites and is left out of this calculation. What was measured about G2, Capterra and Clutch is in for those who sell to companies, G2, Capterra and Clutch appear in some measurements and vanish from others.
How to test whether your company appears in ChatGPT in “X or Y” and in “alternative to X”
To test, ask the AI models your buyer’s comparison questions and note whether your company appears, and in what position.
- Write the questions in four types: “best [category] for [profile]”, “[your product] or [competitor]”, “alternative to [competitor]” and “how much does [category] cost”. The full list is built as in how to build the list of questions to measure brand presence in ChatGPT.
- Ask each question in ChatGPT, Claude, Perplexity and Google AIO, in a new conversation with no history, and more than once, because the same AI answers differently each time.
- Note four things per answer: whether the company appears, in what position in the list, what the AI says about the product and which sites it read.
- Compare your presence with that of each competitor. Presence is how many answers the brand appears in, divided by the total number of answers.
- Repeat the same questions at least once a month, and after each page published. Why a month is the minimum pace is explained in how often to measure brand presence in AI models.
The presence spreadsheet is in how to measure brand presence in ChatGPT.
When the AI gets your product wrong in the comparison
When the comparison brings a plan that does not exist or an old price, the correction starts with a page on your site with the right information. The step-by-step guide is in how to correct wrong information that ChatGPT says about your company, and the case study of the competitor appearing instead of you, in when the AI cites the competitor and not your company.
murmur works for the company to appear in the comparison, and at Visio the search for an alternative opened the door
murmur works for the company to appear in comparison answers, with the same person handling the brand from the diagnostic onward. That person studies the client’s sector, and the guides murmur writes answer that client’s buyer’s questions, on a subdomain of the client’s brand. Week after week, the same set of questions goes to ChatGPT, Claude, Perplexity and Google AIO, and the dashboard, included in the service, shows the company’s presence and that of each competitor. In a software market, it is this ongoing monitoring that shows in which “melhor para” (“best for”) and “alternativa a” (“alternative to”) questions the competitor still appears alone. When the AI gets the product’s plan or price wrong in the comparison, murmur helps correct the information.
Among murmur’s clients that sell software is Visio, operations software for retail chains, franchises and food service, with the same name as Microsoft’s diagramming program. From June to July 2026, in the same 135 questions, Google AIO answers with Visio went from 0% to 45%. In the same period, ChatGPT went from 2% to 22% of answers with Visio, and Claude, from 1% to 7%. The case study page says which questions changed:
As perguntas que destravaram são as de quem ainda não conhece a Visio. São buscas por alternativas e concorrentes de outros sistemas, e perguntas gerais sobre gestão para varejo.
Original quote in Portuguese, preserved as published. In English: “The questions that unlocked it are those from people who do not know Visio yet. They are searches for alternatives to and competitors of other systems, and general questions about management for retail.”
For those who sell software, the search for an alternative is the comparison made by someone who does not know the product yet. On average across murmur’s clients, significant improvement came from 60 days onward, depending on the starting point and the competition. The reasons a page takes more or less time are in how long it takes to appear in ChatGPT.
Frequently asked questions
How does the AI choose the software in a comparison?
The AI chooses the software by what it reads at the moment of the question. In three of murmur’s software studies, vendor sites were among the most read. What other sites say about the company also counts, and whether the AI recognizes its name, as the guide why the AI recommends some companies and not others explains.
Do I need a page from my company against the competitor?
The page against the competitor is worth it when the buyer is looking for that comparison and it brings facts they cannot find anywhere else. The facts about the competitor come from its own site, with a date. One page per competitor with the same text can fall into what Google calls scaled content abuse, which is many pages made to manipulate search, and not to help the reader.
Is being on G2 and Capterra necessary to appear in ChatGPT?
Being on G2 and Capterra is not a requirement, and the weight depends on what the company sells and on the buyer’s question. In murmur’s studies of software for real estate agencies and of BPM, the most-read software comparison site appeared in fewer answers than the most-read vendor site. The weight of each review site is in avaliações e Reclame Aqui influenciam a IA.
Does the “best” list on my site help me appear in ChatGPT?
The “best” list on your site is one of the pages the AI can read at the time of the comparison, and it serves the buyer when it works as a guide to choosing: it compares the options by what each one publishes and puts your product under the same criteria as the others.
Does the buyer trust the comparison the AI puts together?
Software buyers use AI to build the shortlist and check before deciding. The numbers from those surveys, with who measured each one, are in buyers trust the AI recommendation, and check before deciding. The point in the buying journey when the buyer asks AI is in how the B2B buyer finds a supplier through AI.
Where to start
Start with the comparison question your buyer asks before talking to the sales team, and see who the AI lists today. For the first snapshot, all it takes is the site URL in murmur’s free diagnostic, which shows in less than two minutes, without signing up, what the AI models answer in your sector. murmur’s software studies show, sector by sector, which systems the AI cited in a September 2026 round and which sites it read. Each one publishes the 25 questions, which serve as a model for the ones in your sector.
Read also
- How to do GEO for a B2B company: the steps for your company to appear in ChatGPT
- How to do GEO for startups: the steps to appear in ChatGPT against the leader
- GEO for fintechs and banks: fees, timelines and authorization on the site to appear in ChatGPT
- How to get your company recommended by AI, not just cited
- Manipulating what ChatGPT recommends: it worked in a test and puts your site at risk
References
- Lily Ray, “Why Calling Yourself the ‘Best’ Could Be Helping Your Competitors Win in AI Search”, 17 June 2026. There are 100 business “best [category] software” queries, in English, with the AI Overview answers taken from Brand Radar, an Ahrefs tool, on 15 April, 15 May and 8 June 2026; 80 returned an AI Overview (“about 1 in 5 ‘best software’ queries didn’t surface an AI Overview at all”). The author provides SEO consulting and writes “I believe” and “appears” in the readings. https://lilyraynyc.substack.com/p/why-calling-yourself-the-best-could (accessed on 12 October 2026).
- Google Search Central, “Políticas de spam para a Pesquisa Google na Web” (“Spam policies for Google Web Search”), last updated on 16 September 2026, the section on scaled content abuse. https://developers.google.com/search/docs/essentials/spam-policies?hl=pt-br (accessed on 12 October 2026).
- murmur, September 2026 sector studies, with 25 questions without a company name, asked once in each AI, and a screenshot of each answer: legal software, 99 answers in ChatGPT, Claude, Google AIO and Perplexity, https://murmur.marketing/estudos/software-juridico-2026-09/ ; system for real estate agencies, 75 answers, https://murmur.marketing/estudos/software-imobiliaria-2026-09/ ; BPM, 75, https://murmur.marketing/estudos/bpm-2026-09/ ; payroll, 75, https://murmur.marketing/estudos/folha-pagamento-2026-09/ ; time and attendance, 73, https://murmur.marketing/estudos/ponto-eletronico-2026-09/ ; contract management, 75, https://murmur.marketing/estudos/gestao-contratos-2026-09/ ; marketing automation, 75, https://murmur.marketing/estudos/automacao-marketing-2026-09/ ; bidding software, 74, https://murmur.marketing/estudos/software-licitacao-2026-09/ ; dialer and call center, 68, https://murmur.marketing/estudos/discador-call-center-2026-09/ ; customer experience platform, 75, https://murmur.marketing/estudos/plataforma-cx-2026-09/ . Apart from the legal software one, the questions went to ChatGPT, Claude and Google AIO. The sites consulted appear only in the answers that show them. In the nine studies apart from the legal software one, the list is of the Brazilian software that the AI cited at least once; in the legal software one, the table lists all the systems cited. In the how-to-choose and how-much-it-costs questions, in eight studies, there were 15 answers per sector, and 14 in the dialer one, with 1 software from the list in marketing automation, 4 in dialer, 5 in time and attendance and in customer experience, 6 in payroll and 9 in contract management, in bidding and in real estate (accessed on 12 October 2026).
- murmur, Visio case study: from June to July 2026, in the same 135 questions, the answers with Visio went from 0% to 45% in Google AIO, from 2% to 22% in ChatGPT and from 1% to 7% in Claude. https://murmur.marketing/en/cases/visio/ (accessed on 12 October 2026).
- murmur, free diagnostic: https://murmur.marketing/en/diagnostic/ ; studies by sector: https://murmur.marketing/estudos/ (accessed on 12 October 2026).