Security buyers ask AI which vendor to shortlist. What should a cybersecurity company publish?
AI answers about security vendors lean on the vendors' own sites. The pages a cybersecurity company should publish, and the facts each must state plainly.

The short answer
A cybersecurity company should publish, in plain text, the facts security buyers ask AI about before a shortlist: what the product covers, where data is stored, which certifications apply and for what period, dated independent test results, a public advisory record and how its AI features are controlled. AI answers about security vendors draw mostly on vendors' own sites.
Key takeaways
- In Profound's data on 11.84 billion AI citations from 16 April to 16 July 2026, cybersecurity brands drew a median 74% of their citations from their own sites, the highest share of any industry it reported.
- In G2's survey of 1,076 B2B decision makers in March 2026, 51% said they start software research in an AI chatbot more often than in Google, and 64% said they meet inaccurate AI recommendations often or very often.
- Security buyers ask AI specific questions, about coverage, data location, certifications, test results, past incidents and AI controls, and each question needs one page that answers it in its first lines.
- In TrustRadius's survey of 1,862 technology buyers in January 2026, 83% shortlisted three or fewer products, so a vendor missing from the AI answer is usually missing from the evaluation.
- Old advisories, breach coverage and product names left over from acquisitions keep appearing in AI answers until a vendor publishes its own dated, current account.
In this article
What do security buyers ask AI before they call a vendor?
Security buyers ask AI the questions they used to put to analysts, peers and vendor websites: which tools fit a given environment, how two products compare, where data goes, and whether a vendor has had problems. The answer decides which names reach the evaluation. A vendor the engine leaves out has to win its way back in through a referral or an RFP.
The broad B2B evidence points one way. G2, which runs a software review marketplace, surveyed 1,076 B2B decision makers in North America, EMEA and APAC in March 2026. It found 51% start software research in an AI chatbot more often than in Google.2 In the same survey, 69% said AI chatbots had surfaced information that led them to choose a different vendor than they expected.10
Security teams appear in that picture too. The Ponemon Institute and NOLA Marketing published a 2026 study of enterprise cybersecurity decision makers on how they evaluate vendors, including how they use AI overviews and chatbots during selection.9 It covers marketing content and buyer trust broadly. What we add here is the page-by-page work a vendor can do on its own site, tied to what the engines cite.
This playbook suits a security vendor of any size that sells to businesses: endpoint, identity, cloud, email, managed detection or consulting. The examples rotate across buyers in the US, Germany, India and Singapore, because the questions change with the regulation each buyer answers to.
Why your own site matters more in security than in most industries
Your own site matters more because, in security, the engines quote vendors directly. Profound, which sells AI visibility software, analysed 11.84 billion citations across eight engines from 16 April to 16 July 2026. Cybersecurity brands drew a median 74% of their citations from their own sites, the highest of any industry it reported.1 Government and nonprofit organisations sat lowest, at 15.9%.1
Engines differ in how much they lean on brand sites. In the same Profound data, 69% of Gemini's citations went to brand sites, against 47% of ChatGPT's.1 They also differ in how many sources they use. In Semrush's 2026 AI Visibility Index of 126 million US prompts from January to April 2026, ChatGPT cited about 15 sources per response and Gemini about 3.6
AI citations in 2026
In security, the vendor's own site is the main source
Our view: that 74% is good news and a warning. It means a vendor controls most of what the engines say about it. It also means a vague, gated or outdated site is the main reason a security product is described badly.1 Our synthesis of citation sources by industry sets the security figure next to other sectors.
Who in the security buying group asks the engines, and what?
Several people ask, and each asks a different question. Forrester's State of Business Buying 2026, fielded in 2025 with exact dates not stated, found buying decisions involve on average 13 internal stakeholders and 9 external participants.8 In a security purchase, that group usually spans the security lead, IT operations, procurement, legal and privacy, and often an outside consultant or managed service provider.
Each of them puts a narrow question to an engine. The security architect asks what the product detects and on which platforms. The privacy lead asks where telemetry is stored and for how long. Procurement asks how pricing works. The consultant asks how the product compares with the one already in place. Every question maps to a page, and every missing page is a question the engine answers from someone else's site.
| Who asks | Typical question to an AI engine | Page that should answer it |
|---|---|---|
| Security architect | Does it cover our operating systems and clouds? | Coverage and integrations |
| Privacy and legal | Where is our data stored, and for how long? | Data handling and residency |
| Risk and compliance | Which certifications does the vendor hold? | Trust centre summary |
| Security lead | How did it do in independent tests? | Test results, dated |
| Everyone, at the fact-check | Has this vendor had a breach? | Advisories and incident record |
| Procurement | How is it priced? | Pricing structure |
| CISO | How are the product's AI features controlled? | AI features and controls |
Our page map for a security vendor. Illustrative; adjust the roles to your own buying group.
The seven pages a security vendor should publish in plain text
Seven pages carry most of the questions above. Each should open with the answer in a sentence an engine can quote, carry a date, and match what your sales engineers say on calls. The common thread is simple: put facts in HTML on an open page, and keep the detail you protect behind a form.
Our general guide to the seven pages a SaaS site must get right covers pricing, comparisons and reviews for any software company. The list below is the security version, where the stakes of a wrong fact are higher.
1. Coverage and integrations
State what the product protects and where it runs: operating systems and versions, cloud platforms, identity providers, email systems and the SIEM or ticketing tools it sends data to. Say what is not supported as clearly as what is. An engine asked "does this work on our Linux servers" needs a sentence, not a logo wall.
2. Data handling and residency
Say what data the product collects, where it is processed and stored, which hosting regions a customer can choose, and how long data is kept by default. A buyer in Germany or India often has to answer a data-location question before anything else, and the engine will answer it for them from whatever page states the facts.
3. Certifications and attestations
List each certification with its scope and dates: the standard, which products and locations it covers, the audit period or certificate dates, and how to request the full report. Keep the reports gated if you must. The summary should be open, because an engine cannot quote a document behind a request form.
4. Independent test results
Publish the results of independent tests your product took part in, with the test name, the date, the product version tested and a link to the tester's own page. Include the tests where you did less well, or say which ones you skipped. Never write "#1" or "best" from a single test; buyers check, and engines repeat claims they cannot verify.
5. Advisories and incident record
Keep a public security advisory page for your own products, with dates, affected versions and fixed versions. If you have had a public incident, publish a dated account of what happened and what changed. The engines will find the coverage either way; your own account is what gives them the current facts.
6. Pricing structure
Say what you charge per: endpoint, user, asset, data volume or a flat platform fee, plus the minimum term and what costs extra. Many security vendors keep prices private. Even then, the unit and the structure can be public, and they answer most of what procurement asks an engine.
7. AI features and the controls around them
Explain what your product's AI features do, what data they use, who can switch them off, and how any automated actions are limited and logged. Security leaders are asking these questions about their own estates. In Gartner's survey of 297 cybersecurity leaders in the second quarter of 2026 (geography not stated), 54% said their organisation had no defined approach to limiting AI agent access, or relied on predefined human access.7 A vendor that answers the question plainly helps the buyer who has not yet answered it internally.

What an engine can quote, and what it skips
Vague
- "Enterprise-grade protection everywhere"
- "Certified to the highest standards"
- "Award-winning detection"
- "Contact us for pricing"
Quotable
- Supported systems and versions, listed
- Standard, scope and audit period, dated
- Test name, date, version and link
- Pricing unit, minimum term and extras
How do old breaches, advisories and acquisitions show up in AI answers?
They show up as if they were current, unless a dated, current account exists next to them. Engines read news coverage, forum threads and archived product pages as well as your site. A three-year-old incident report with no follow-up can become the answer to "is this vendor safe to buy".
Buyers notice the errors. In G2's March 2026 survey, 64% of B2B decision makers said they encounter inaccurate AI recommendations often or very often.3 In security, the inaccuracy that hurts most is usually a stale one: a vulnerability that was fixed, a product line that was retired, or a certification that has since been renewed.
Acquisitions add a second problem. Security vendors buy and merge often, and the engines may keep naming the old product, the old company or a discontinued plan for months. If that has happened to you, our guide to getting AI answers to use a new company name sets out the redirects, profile updates and statements that help.
Our view: do not hide an incident page to keep it out of the answers. Removing your own account leaves only other people's versions. A short, dated page that states what happened, which versions were affected and what changed gives the engine a current source, and gives your sales team one page to send. For wrong facts that are not about incidents, our guide to fixing what ChatGPT gets wrong about your company covers tracing and correcting them.
Example prompts for buyers in four countries
Write test prompts the way your buyers ask, with their country, regulation and environment in the question. These are illustrative, written for a fictional endpoint security vendor, Fabrikam Security.
- US: "Which endpoint security tools support macOS and Linux servers and integrate with our SIEM, for a 2,000-employee company?"
- Germany: "Which managed detection providers can store our data inside the EU, and which certifications do they hold?"
- India: "What endpoint protection do mid-sized Indian banks use, and how is it priced per device?"
- Singapore: "Has Fabrikam Security had a breach, and is it safe for a regulated financial firm in Singapore?"
- Any market: "How does Fabrikam Security control the automated actions its AI assistant can take?"
Only two of them name the vendor. The unbranded prompts decide whether you reach the shortlist; the branded ones decide whether you survive the fact-check. Run both kinds on ChatGPT, Gemini, Microsoft Copilot, Google AI Overviews and Google AI Mode, on more than one day, because answers change from run to run.
Do security buyers trust what AI tells them?
Partly, and they check it with people. Gartner surveyed 645 B2B buyers in August and September 2025 (geography not stated) and found 69% prefer to validate AI-generated insights with sales reps.5 In the same survey, 51% said they were more likely to meet misleading information from generative AI, and 49% from a sales rep.5
The checking is routine, and the shortlist is short. TrustRadius, a software review site, surveyed 1,862 technology buyers in January 2026. It found 83% shortlisted three or fewer products, and 94% of the buyers who used AI said they fact-check its answers at least some of the time.4
Our view: in security, the fact-check is where deals are lost. A buyer arrives at the first call with an AI answer about your coverage, your certifications and your last incident. If your sales engineer, your website and the engine give three versions, the buyer has a reason to doubt all of them. Make the website the version of record, and brief sales from it.
How buyers handle what AI tells them
Which page should a security vendor fix first?
Fix first the page behind the question that most often drops you from the answer. Start from what the engines say today, not from the page you like least. Run the prompts, record which ones leave you out or misstate a fact, and trace each gap to the source the engine cited instead.
Three gaps are common on security sites and quick to check: a gated trust centre with no open summary, a coverage page that lists features but not platforms, and an old incident with no current account. Each is a few days of work for people you already employ. Test results and AI controls take longer, because product and legal teams have to agree the wording.
A 30-day plan for a security vendor's pages
Write 60 to 120 prompts across the seven questions, branded and unbranded, for each country you sell in.
Run each prompt on five engines on three separate days; record names, claims and cited sources.
Mark missing or wrong facts and trace each to the page or third-party source the engine quoted.
Open the trust centre summary, rewrite the coverage page, publish or update the incident account, then brief sales.
At Sigzen AI, our AI answer monitor runs prompt sets like this across ChatGPT, Gemini, Microsoft Copilot, Google AI Overviews and Google AI Mode, through a licensed AI-answer data provider. Our AI Visibility Audit runs 150 buyer prompts on three separate days, against five competitors, and maps the sources the engines trust in your category.
What these pages will not fix
Clear pages make you quotable. They do not make you the obvious choice in a category where a rival has years of analyst coverage, test history and community discussion. Engines weigh what others say about you as well as what you say, and that takes longer to change than a page.
The figures here have limits too. Profound reports industries as examples rather than a full table, and its figures are medians across brands, so your own share may sit far from 74%.1 The buyer surveys cover B2B and technology buyers in general, not security teams alone, and several come from companies that sell into software buying or search marketing. Read them as direction, not precision.
Avoid the shortcut of self-ranked "best security tools" lists on your own blog. Buyers in security are trained sceptics, and our review of the evidence on self-ranked lists explains why they are a weak bet for AI recommendations as well.
Our view: security is the industry where a vendor's own pages carry the most weight in AI answers, and most security sites still lock their best facts in PDFs, forms and sales calls. Open the summary, date every claim and keep one current account of your history. That is the part of AI visibility a security vendor fully controls.
Sources
- Profound, where AI citations come from: 11.84B citations, 8 engines, 16 Apr–16 Jul 2026 (Jul 2026)
- G2, 2026 AI Search Insight Report: 1,076 B2B decision makers, North America, EMEA and APAC, Mar 2026
- G2, 2026 AI Search Insight Report: 1,076 B2B decision makers, North America, EMEA and APAC, Mar 2026
- TrustRadius, Beyond the hype: 1,862 technology buyers, Jan 2026 (Jul 2026); 94% is of buyers who used AI
- Gartner, survey of 645 B2B buyers, Aug–Sep 2025; geography not stated (May 2026)(dated)
- Semrush, 2026 AI Visibility Index: 126M US prompts, Jan–Apr 2026 (Jun 2026)
- Gartner, survey of 297 cybersecurity leaders, Q2 2026; geography not stated (Sep 2026)
- Forrester, State of Business Buying 2026: buying groups, fielded 2025 (Jan 2026)
- Ponemon Institute and NOLA Marketing, The State of Cybersecurity Marketing Influence 2026 (Jun 2026)
- G2, 2026 AI Search Insight Report: 1,076 B2B decision makers, Mar 2026


