Why Isn't My Product Showing Up in Claude, Gemini or Perplexity?

Your product can appear in Perplexity and still be absent from Claude or Gemini because those experiences do not use the web in the same way on every question. The assistant may not search, a crawler or firewall may block your page, or there may be too little independent information to support a recommendation. Check which problem you have before changing your site. No setting can force a mention.

We run LaunchAF, which offers a free GEO checker, so weigh that accordingly. This guide focuses on the differences among the three services. For the general reasons a new brand can be missing, read our ChatGPT visibility diagnosis rather than repeating its nine fixes here.

Key takeaways

• Test a brand lookup and an unbranded buyer question separately. Recognition is not the same as recommendation.
• Claude can answer without a current web search; its search and user-fetch crawlers have different jobs from its training crawler.
• Google Search's AI features and the Gemini app are related but not identical. Googlebot access and Search indexing are essential checks for Search, while Google-Extended is a separate Gemini control.
• Perplexity normally shows numbered source links. Inspect what it cites, not just whether it names you.
• A reachable, accurate product page and a few relevant independent references help people verify you. They do not guarantee an AI answer.

Vendor documentation and interface behavior below were checked on October 14, 2026. Features can differ by plan, region, workspace settings, and product version. Treat the table as a way to choose your first diagnostic check, not as a map of any vendor's private source-selection system.

Do Claude, Gemini and Perplexity find products the same way?

No. An assistant can use model knowledge, retrieve current material, or combine both. The vendor's documented controls tell you what can be crawled or fetched, but they do not reveal why a particular answer chooses one product over another.

Anthropic documents the three Claude agents on its crawler-support page. Google describes Search AI eligibility separately from Google-Extended, and its Gemini source help says not every response has source links. Perplexity documents its crawler roles and numbered citations. These are documented capabilities, not evidence that your page was fetched for a given answer.

Why isn't my product showing up in Claude?

First ask whether Claude searched the web for your question. Anthropic's web-search help says Claude uses search for questions that benefit from current information and cites sources in web-search responses. Some interfaces let a user toggle search, while Anthropic says its newer experience may search automatically when useful. Workspace settings and usage limits can also matter. If the answer contains no current sources, do not interpret a missing mention as a failed crawl. Try the same buyer question with web search available and note what happened.

Next, separate three crawler names that are easy to confuse. Anthropic says Claude-SearchBot helps its search responses, Claude-User can fetch a page following a user request, and ClaudeBot collects material that could contribute to future model training. A founder may choose to block training while allowing search and user-directed retrieval. The choice is yours; there is no requirement to open private pages. Read the rules for the relevant public URL, not only the root of robots.txt.

A permissive robots file is only one layer. A CDN challenge, WAF rule, login wall, noindex, or a page that returns an empty shell to simple requests can still make a product hard to use as a source. Check server or firewall logs if you have them. Our robots.txt tester evaluates a URL against rules you provide; it cannot prove that Anthropic's infrastructure successfully fetched the page. The free GEO checker adds a broader access and page-content audit, but its sample firewall probes are not a guarantee that Claude-SearchBot passed your CDN. Look for that agent in your own logs if you need confirmation.

Finally, consider identity. A young brand with a common name may be recognized only when you include its domain. Try “What does [Brand] at [domain] do?” and then an unbranded question your buyer would ask. If the first works and the second does not, your main issue may be relevance or independent context, not access. Keep the category and use case clear on your own site; seek accurate descriptions from real partners or publications rather than manufacturing mentions. The AI citation guide covers the general content work without claiming a Claude-specific shortcut.

How do I get my SaaS mentioned in Gemini?

Start by naming the product surface. A Google AI Overview or AI Mode result inside Search is not the same as a response from the Gemini app. Google's Search Central guidance says pages supporting its Search AI features must be indexed and eligible to appear with a snippet. It says ordinary SEO fundamentals apply and no special AI file or schema is required. An indexed page is eligible, not promised a supporting link.

Use Google Search Console URL Inspection for the exact product page. Check whether Google can index it, which canonical URL Google selected, and whether the page has a snippet restriction. Make the product's name, category, intended user, and main action visible in text. If the homepage claims only to “transform workflows,” a buyer and a search system both have little to work with. An accurate product page, a direct answer to a customer question, and sensible internal links are more defensible than adding unsupported AI markup. Our GEO guide for startups covers those shared foundations in depth.

The Gemini app requires another distinction. Google's crawler documentation says Google-Extended is a control for training future Gemini models and for some grounding uses in Gemini apps. The same document says it does not affect inclusion or ranking in Google Search. Therefore, blocking Google-Extended is not the same as blocking Googlebot, and it does not remove your site from ordinary Search by itself. Nor can we conclude that it removes every possible mention from every Gemini response. Do not change your training or grounding preference merely to chase a brand appearance; understand the trade-off first.

Google also says Gemini app responses may show related source links, but not all do. If you see a wrong brand description with no source, record the prompt and result rather than pretending you know which page caused it. If a linked source is wrong, correct the first-party page or the third-party profile you control. Consistent product facts across your site and relevant independent profiles help readers identify the right company; they do not guarantee a Gemini recommendation.

How do I get Perplexity to cite my website?

Perplexity makes the source check unusually visible. Its help center says answers include numbered citations. Ask a buyer-style question about your category, then open those links. Do the cited pages directly answer the question? Are they product pages, directories, reviews, or articles? Is your own page absent because it is inaccessible, or because it does not answer what the question asks? You cannot infer its exact selection weights, but you can inspect the evidence a reader sees.

Perplexity says PerplexityBot is designed to surface and link sites in search results and is not a foundation-model training crawler. Perplexity-User supports user-triggered visits. Its crawler documentation also notes that robots settings and permitted IP ranges can matter. If you want public pages available, check robots.txt, your WAF, and the actual response code for a public product or answer page. Do not open private account routes just to make a test green.

Then give the system a page worth citing. For “Does [Product] export to CSV?”, answer yes or no near the top, explain constraints, and keep the page current. For “Which tools help with [task]?”, make your product's use case explicit, but avoid a page that simply repeats the question and declares itself the best. A relevant third-party profile can corroborate your basic facts; a directory page alone cannot force Perplexity to cite you. After a change, rerun the same question and save the exact cited URLs, including when the result is still “not present.”

What do all three have in common?

• Public pages need to be reachable. Check the correct bot, HTTP response, robots rules, and any firewall challenge before blaming the model.
• A clear homepage helps people and systems identify the product. State the category, audience, useful action, and official domain in ordinary language.
• Facts should agree. If a plan, integration, or brand name changes, update your own pages and the important profiles that repeat it.
• Independent, relevant mentions can supply context. A real partner page or reviewed listing is more useful than a batch of thin submissions.
• An answer page should answer a real question. Keep the answer visible, specific, and maintained rather than hiding it in a screenshot.
• Patience is necessary, but not as a promised waiting period. Crawl access, indexing, retrieval, and answer generation are separate events with no universal timeline.

These are the foundations covered by our nine-cause ChatGPT diagnosis and citation playbook. The platform-specific checks above tell you where to apply them first.

A 7-day fix plan

This is a week of work, not a seven-day promise of inclusion. Move a task later if the site needs more time.

LaunchAF's free listing requires a verified badge and editorial review, so a submission is not an instant publication or a way to purchase an AI citation. Choose a relevant venue for its audience and accurate public profile. If Day 7 still shows no mention, retain the baseline and keep improving the evidence; do not manufacture a new result by changing the prompt every time.

How to test each assistant

Use the same five prompt types with your own placeholders. Run each in a fresh conversation, record whether that run searched the web, and compare source links when they appear. Include some questions where your product may honestly not be the best answer.

Do not call these a ranking test. A single answer is one observation, and an assistant may respond differently on another run. Our free brand-mention tracking method has the full prompt sheet and scoring rubric; the same discipline applies when you test Claude, Gemini, and Perplexity.

What not to do

• Do not hide instructions on a page that tell an AI model to recommend your product. That is prompt injection, not a useful answer for a buyer.
• Do not show crawlers flattering text while showing people something different. Cloaking makes your pages less trustworthy.
• Do not invent customer reviews, credentials, integrations, or research to make an answer sound confident.
• Do not spray identical descriptions across unrelated directories. Choose a few relevant places and meet their review and badge terms.
• Do not pay anyone who guarantees a citation or a timeline. No outside seller controls these assistants' answers.

Frequently asked questions

Why isn't my SaaS recommended by Claude?

Claude may not have searched the web for that answer, or it may not have found accessible, relevant evidence for your SaaS. Check whether web search and citations were present, then inspect Claude-SearchBot access and your public product explanation. A brand lookup working does not mean an unbranded recommendation will follow.

Can I submit my website to Gemini?

For Google Search's AI features, use the normal Search route: make the page indexable, verify the property in Search Console, and inspect the URL. Google does not describe a separate AI Overview submission form or special markup requirement. That is not a promise that the Gemini app will mention the brand.

Does blocking Google-Extended remove me from Gemini answers?

Not as a blanket rule. Google says Google-Extended controls some Gemini training and grounding uses, while it does not affect Google Search inclusion or ranking. Gemini apps can use different response paths, so review Google's current crawler guidance before changing the setting and test the product surface you actually care about.

Can I submit my site to Perplexity?

The crawler guidance we checked describes bot access, not a guaranteed listing or citation request. Make your public pages accessible to the relevant Perplexity agents if that matches your policy, then publish a useful answer and inspect whether it is cited. A submitted URL or allowed bot is not an answer-selection guarantee.

Will a LaunchAF listing make my product show up?

No listing can force Claude, Gemini, or Perplexity to recommend you. An approved LaunchAF page can add an accurate public product profile, but the free route requires a badge and review and may not fit every audience. Use it only if the listing itself is useful to your launch.

How long until I appear?

There is no reliable universal timeline. Crawling, indexing, retrieval, and the answer vary by product and prompt. Keep a dated record of access, source links, and repeated tests so you can tell a real improvement from one fortunate response.

The bottom line

Do not troubleshoot three assistants as if they share one index. Identify the product surface and whether it used the web, check the appropriate public-page access, and read the sources it actually shows. Run the free GEO checker, then fix the first failing item you can verify. Retest the same questions and judge the pattern, not one answer.