How to Get Your Cybersecurity Product Recommended by ChatGPT, Perplexity and Claude: A 10-Category Test (2026)

I typed one question into ChatGPT: “best SIEM software in the US.” It named ten vendors. Microsoft Sentinel came first. Splunk and CrowdStrike followed. Then I asked Claude the same thing, and it put Splunk first. Google’s AI Overview agreed with Claude. Perplexity sided with ChatGPT. Four engines gave two different winners. And none of them asked for a demo, a form or an email address. They simply told me who to call.

That moment now decides deals. G2 surveyed 1,076 software buyers in March 2026. Half of them (51%) start with an AI chatbot. Even more telling, 69% ended up buying from a different vendor than the one they first planned to buy. So the shortlist forms before anyone visits your site. In security, the stakes are heavier still. IBM puts the average data breach at $4.99 million, a record high. Verizon says 31% of breaches now begin with a software vulnerability. Gartner expects about $244 billion in security spending this year. In short, buyers have the money and the fear. They ask AI first, and they act on what it says.

So I tested it properly. I asked ChatGPT and Perplexity ten questions that security teams type every week. Then I ran the same ten searches on Google. The result is uncomfortable. ChatGPT named 88 vendors, yet only 21 had their own page on Google’s first page. Microsoft appeared in six of the ten answers. And in four categories, the two engines could not agree on who comes first. Ranking on Google, it turns out, does not put you on the shortlist. Something else does. This article shows you what it is.

This article is for the people who sign off on security budgets and security growth: founders, CMOs, heads of demand generation, and the CISOs who want to understand how their shortlists are being shaped. It will show you what the engines say, where they get it, and what to publish so your product gets named. You will get:

• The data: ten categories, 88 named vendors, two AI engines and Google side by side.
• The screenshots: what a buyer actually sees in ChatGPT, Perplexity, Claude and Google.
• Seven findings: including which sources each engine trusts, and one surprise about vendor-written roundups.
• The access check: why your CDN or firewall may be hiding you from AI.
• The plan: ten steps, a trust stack, four page templates and a 12-prompt test.

If you want the plan first, jump to it. Otherwise, start with the test.

Why security buyers ask AI first, and why it matters to your pipeline

Three forces are pushing in the same direction. First, buying is moving to chatbots. Second, security is among the sectors where AI already sends the most visits. Third, security buyers verify before they trust, and AI shortlists shape who gets verified.

Read those together. Security buyers are afraid, well funded and early adopters of AI. That is the ideal audience for an engine that answers “who should I buy?” in ten seconds. Sources for each number: G2, Conductor, IBM, Verizon and Gartner. For wider context, see my AI search statistics.

The test: 10 security buying questions on ChatGPT, Perplexity and Google

I picked ten categories that security teams buy every year. Each is a question a real buyer would type. For ChatGPT and Perplexity, I added “in the US” so the answers would not default to my location in Kigali. For Google, I ran the same words with US settings and personalization off. Then I recorded three things. First, which vendors ChatGPT named, in order. Second, which vendor Perplexity led with. Third, which of ChatGPT’s vendors had their own website among Google’s page-one organic results. Reviews, forums and analyst pages did not count as a vendor’s own page.

What a security buyer actually sees

Numbers help. Screenshots show the experience. Start with ChatGPT. Here is the answer to “best endpoint protection software for small business in the US.” Notice the structure. It opens with one recommendation. It ranks eight products. And each product gets a short “best for” tag, four bullets and a link.

Then comes the part most marketers have not seen. ChatGPT did not stop at the list. It added a short intake form at the bottom of the answer, with four questions: how large is the business, which devices need protection, what is the monthly budget per device, and does the company have an IT or security team. It even pre-selected likely answers.

That intake form is a warning and a gift. It is a warning because ChatGPT now qualifies the buyer on your behalf: size, devices, budget, team. If your pages never mention those four things, the AI has nothing to match you against. It is a gift because you now know what to publish. State who you are for, in those exact terms.

Now compare Google. For “best SIEM software,” Google put an AI Overview on top. It led with Splunk Enterprise. Next to it, it showed a Reddit thread and a Gartner page.

Scroll down and the classic results tell a similar story. The first result was a “best SIEM tools” roundup written by a vendor, Palo Alto Networks. After it came Reddit, Gartner, and another vendor roundup, this one from SentinelOne.

Perplexity gave a third answer. It led with Microsoft Sentinel, like ChatGPT, and it showed its sources in a panel on the right. Ten sources were cited. Four of them were vendor-written “top 10” pages, from Palo Alto Networks, Huntress, ManageEngine and SentinelOne.

Claude took a fourth path. It searched the web, then led with Splunk. Its stated reason was analyst recognition: Splunk “was named a Leader in the 2025 Gartner Magic Quadrant for SIEM.” And many of the pages it cited were the vendors’ own announcements of that result, from Splunk, Google Cloud, Exabeam and Gurucul.

Four engines gave two different leaders. That is the whole story in one table. A buyer who asks one engine sees one world. A buyer who asks two sees a conflict.

7 findings that should change your plan

1. Google rank does not decide who gets named

Only 21 of the 88 vendors ChatGPT named (24%) had their own page on Google’s first page. In vulnerability management and identity, the number was zero. Okta, CrowdStrike, Qualys and Tenable were all named by AI without a page of their own on page one for that search. So you can rank and not be named. You can also be named without ranking. The two systems pull from different shelves.

2. Engines agree on market leaders and split in crowded markets

ChatGPT and Perplexity gave the same first pick in six categories: SIEM, vulnerability management, identity, SOC 2 automation, awareness training and cloud posture. In four they split: endpoint protection, penetration testing, MDR and email security. Notice the pattern. The six agreement categories have a clear category leader or two. The four split categories are crowded, and many vendors look alike. In a crowded market, small signals decide, and different engines weigh them differently. That is where an unknown vendor has the best chance to win a spot.

3. Suites win by repetition

Microsoft appeared in six of ChatGPT’s ten answers: Defender for Business, Sentinel, Defender Vulnerability Management, Entra ID, Defender for Office 365 and Defender for Cloud. CrowdStrike appeared in five. Rapid7 and Huntress each appeared in four. SentinelOne appeared in three. A vendor with a product in many categories gets many chances to be named, and each mention reinforces the others. If you sell several products, link them clearly and publish the same facts, in the same words, for each.

4. Each engine trusts a different shelf

This matters more than any single ranking. If you invest only in Google, you miss Perplexity’s small roundup sites. If you invest only in analyst relations, you miss Reddit. Spread your proof across the shelves each engine reads.

5. Vendor-written roundups are everywhere, and they are a risk

SentinelOne’s own “best of” pages held page-one slots on six of the ten Google searches. Palo Alto Networks, Huntress, Orca, Rapid7, Synack and Defendify did the same in their categories. And Perplexity cited them. It is a smart tactic: the buyer asked for the best, and the vendor wrote the list. But it has limits. ChatGPT named SentinelOne in only three categories, and it never named several vendors that rank well on Google, such as Exabeam, Sublime Security, Defendify and Scytale. A roundup that puts your product first and gives no honest trade-offs may win a Google slot and still lose the buyer’s trust. If you publish one, disclose your interest, name rivals fairly and date every fact.

6. ChatGPT is turning answers into intake forms

The endpoint answer ended with a four-question form: company size, devices, budget per device, and IT team. Buyers will increasingly answer those questions and get a narrower list. So your pages should say, in plain words, which company size you serve, which platforms you support, what you cost per seat or per device, and whether you need an in-house team to run the product. Vague “for every business” positioning gets filtered out.

7. Ranking is not recommendation

Several vendors earned Google page-one slots for their roundups and were still not named by ChatGPT. That echoes my findings on AI citation versus AI recommendation. Being cited is one thing. Being recommended is another. Citations come from pages that answer the question. Recommendations come from consensus: many sources that agree you belong on the list.

3 questions for your next leadership meeting

If you run a security company, take these three questions to your leadership team this week. They take an hour to answer, and they will tell you whether AI is helping your pipeline or quietly draining it.

Then do a back-of-envelope check on what silence costs you. Multiply four numbers: the qualified opportunities you expect this quarter, the share of those buyers who start in an AI chatbot, the share of those chats where you are not named, and your average contract value. G2’s survey puts the first-step share at 51% for software buyers overall. Your own mix will differ, so use your own sales data. If the product of those numbers makes you wince, you have found your budget line.

What decision makers ask AI, and what vendors should publish in return

Security buyers do not ask “what is the best product.” They ask narrower, riskier questions. Here are the questions I would expect a CISO, a head of IT or a founder to type, and the page that should answer each one.

Notice the last row. Breaches drive searches, and most vendors publish nothing within the first day. I used the same approach in my Binance hack study. A short, sourced explainer with timestamps and an update log gets quoted when no one else has published.

How an AI shortlist forms

Put the findings together and a pattern appears. The buyer asks. The engine reads a footprint of public sources. It weighs fit, recognition, proof and consensus. Then it names a few products and asks the buyer to narrow.

Every box in that flow is something you can influence. The question is shaped by your content. The sources are shaped by your PR, reviews and pages. The weights are shaped by your proof. And the final answer is shaped by how clearly you state who you serve.

Open the door first: your firewall may be hiding you from AI

Here is a risk unique to security companies. You probably sit behind a content delivery network and a web application firewall. That is good practice. But it can also block the bots that feed AI search. On July 1, 2025, Cloudflare changed its default so that AI crawlers are blocked unless a site owner allows them, and every new domain is asked at sign-up whether to allow them. Other providers have their own bot rules. If your marketing site has a strict bot policy, check it.

The good news is that Google needs nothing special. Google says on its AI features page that “there are no additional requirements to appear in AI Overviews or AI Mode.” The other engines use search bots you can allow separately from the bots that collect training data.

Test with your own tools. Request a key product page as each bot and confirm it returns the full page, not a challenge screen. For setup details, read my AI crawler guide. Bot facts come from OpenAI, Anthropic and Perplexity.

The security vendor trust stack

Security is a trust business. So the order of work matters. Start at the bottom, because nothing above it works if the engines cannot read you.

A 10-step plan to get your security product named

Steps 1 to 4 build the base. Steps 5 to 8 win the shortlist. Steps 9 and 10 keep you honest and current.

For schema, add Organization sitewide and Article and FAQPage on guides, matched to the visible page. My schema guide shows safe patterns. And keep one rule above all: claim only what you can prove. “Stops all ransomware” is a claim that a CISO, a journalist or an AI engine can check and quote against you.

4 page templates security buyers read

Template 1: the “best for” comparison page

Template 2: the trust page

Template 3: the “versus” page

Template 4: the incident explainer

The 12-prompt test: see where you stand today

Run each prompt on ChatGPT with web search on, then on Perplexity, Claude and Google. Run it three times, because answers vary. Record whether you are named, your position, what the engine says and which sources it shows.

If you want it automated, see my AI visibility tools review. For the manual method, read how to check what ChatGPT says about your business. And for the same method in another category, see my SaaS “best software” test and the fintech study.

Where I come in

I have spent six years writing for crypto and finance outlets, where a wrong fact costs readers money. Now I test what AI engines say about brands. For a security company, the work looks like this:

I cannot promise a ranking or an AI citation. I can promise a clear fix list, sourced data and honest reporting. Prices come from my pricing page, checked November 4, 2026.

How I checked, and the limits

• One snapshot. I ran ten ChatGPT prompts, ten Perplexity prompts and ten Google searches on November 4, 2026. Answers vary run to run and by user. Google varies by day and location.
• Claude and Google AI Overview on one question only. I tested them on SIEM. Do not read their SIEM results as their behavior across all ten categories.
• Personalization. ChatGPT and Claude were signed in to my accounts and Perplexity to a free account. I added “in the US” to each prompt, yet account memory may have shaped some answers.
• Wording. The Google searches used the same words as the prompts, without “in the US,” with US settings applied.
• “Own page” is my judgment. A vendor counted only when its own domain appeared in the page-one organic results. Reddit, Gartner, review sites and videos did not count. Vendor-written roundups did count as the vendor’s own page.
• “First” means first named. ChatGPT lists vendors in an order, and Perplexity opens with a summary sentence. Neither always calls one vendor the winner. For Perplexity, I used the first vendor named in the opening answer.
• Sources are partly visible. ChatGPT showed source names in four of ten answers. I counted Perplexity’s source panel for all ten, which gave 94 citations.
• The intake form. I saw ChatGPT’s four-question form in the endpoint answer. I did not check the other nine.
• Third-party data has its own limits. Conductor covers US enterprise sites. G2 surveyed buyers who chose to respond. IBM, Verizon and Gartner figures are read from their own pages. Check the originals before you reuse a number.

Related reading: SEO, AEO and GEO explained and why AI engines rarely cite CoinGecko.