LightSite AI Review: Tracking AI Visibility, Mentions and Structured Data

What does LightSite do?

LightSite describes four layers, run by specialised agents:

• Infrastructure. It deploys and maintains a machine-readable layer on your domain. That layer includes JSON-LD entities for your business, an AI-specific sitemap, and structured endpoints such as a company profile, products and FAQs. It also adds an agent manifest and Skills API endpoints that AI agents can call.
• Measurement. It tracks AI bot crawling, extraction success and failure, Skills API calls, and changes in citation share over time.
• Competitive intelligence. It scans the main assistants for your brand's presence, sentiment and citation share, and shows query-level reasons competitors are cited.
• Execution. Agents update structured data, draft content for visibility gaps, find backlink opportunities, run outreach and ship prioritised fixes.

It also offers free diagnostics: a website analyser, a check of whether AI crawlers can reach your pages, a mentions tracker and a content quality scorer.

LightSite says it targets mid-market B2B software, learning and HR technology brands. It also names lean teams without developer time, and brands with generic names that assistants confuse with other companies.

How much does it cost?

LightSite's Shopify app lists an Annual Pro Plan at $3,588 a year, about $299 a month. It includes advanced analytics, competitive insights, custom prompts and a seven-day free trial. The app launched on the Shopify App Store in August 2025. LightSite's pricing page lists Foundation at $129 a month (500 prompts a month, one user, a basic AI-readable layer, bot analytics and mention tracking) and Pro at $299 a month (2,700 prompts, competitor, sentiment and citation-source tracking, and full structured data implementation), both billed annually with a seven-day free trial. Enterprise is quoted by traffic, domains and users. The Shopify app's $3,588 a year matches Pro. Ask how many pages the machine-readable layer covers and how much agent work is included each month.

Does the machine-readable layer help AI citations?

This is the crucial question, and the independent evidence is weak.

• Google says no special structured data is needed for AI Overviews or AI Mode, and that structured data should match visible content.
• Assistants that fetch pages read the visible text. Tests by SearchVIU found ChatGPT, Claude, Perplexity, Gemini and AI Mode ignored JSON-LD when fetching a page.
• An Ahrefs study of pages that added structured data, compared with matched controls, found no measurable lift in AI citations.
• AI sitemaps, agent manifests and llms.txt-style files have no documented use by the major assistants. Google says it does not use llms.txt.

LightSite's own research reports a 30-day experiment across more than 6 million bot requests. In it, pages with structured endpoints had 12% higher extraction success, 17% greater crawl depth and a 13% higher crawl rate. Those are vendor figures, on metrics the vendor defines. More crawling is not the same as more citation, a gap covered in whether being crawled by an AI bot means your pages will be cited. Customer quotes on its site report gains such as 13% more organic LLM visits in 90 days, but without controls it is hard to separate the layer's effect from the content and outreach work done at the same time.

The fair conclusion is that the layer is probably harmless if it matches your pages, and possibly useful for agents that call its endpoints. There is no independent proof that it raises citations.

What should you watch out for?

• Consistency with visible content. Structured endpoints and JSON-LD must say the same as your human-facing pages. Serving AI systems content that differs from what people see risks being treated as cloaking, as covered in whether serving different content to AI crawlers is cloaking.
• Duplicate content. Separate machine-readable pages that repeat your product or FAQ content should not compete with your main pages in search. Ask how they are handled with canonical tags or noindex.
• Who controls the layer. Find out what happens to the endpoints if you cancel, and whether you can export them.
• Automated outreach and content. Review drafts and emails before they go out. Heavily automated content risks Google's scaled content abuse policy, and outreach must follow disclosure rules if anything is offered in return.
• Measurement on the vendor's own layer. "Extraction success" on LightSite's endpoints is not the same as assistants using your content in answers. Judge results on mentions and citations in the answers themselves.

Which parts are most useful?

• Bot analytics. Knowing which AI crawlers reach which pages, and where they fail, is valuable whatever you think of the layer. You can get similar data from server or CDN logs, but a ready dashboard saves time.
• Mentions and competitor tracking across five major surfaces, with query-level competitor reasons.
• Entity disambiguation for brands with generic names. Consistent, well-structured facts about the company may help systems that do read markup to separate you from similarly named firms, although the visible pages and third-party sources matter more.
• Execution capacity for teams without developers or content staff, if you review what the agents produce.

How does it compare with other tools?

Most trackers measure answers and stop at advice. LightSite's pitch is doing the work, both technical and editorial. That puts it closer to agency-style platforms than to pure trackers. If you mainly need tracking, cheaper self-serve tools exist, as covered in the RankPrompt review and the Semrush AI Visibility Toolkit review. Which special files are actually worth publishing for agents is covered in what should go in a website's /.well-known/ directory for AI agents.

How should you trial it?

1. Record a baseline first: mention rates for 20 to 30 prompts, run several times in each assistant, plus current AI bot activity from your logs.
2. Deploy the layer on a subset of pages if possible, so you can compare against similar pages without it.
3. Keep content and outreach changes separate from the layer during the test, or you will not know which one worked.
4. Compare after four to eight weeks on mentions and citations in answers, not only on crawl metrics.
5. Check the endpoints yourself to confirm they match your visible pages.

What do people ask about LightSite AI?

Does LightSite require code changes?

LightSite says it deploys without a development sprint, and customer quotes describe setup in minutes. Ask exactly how it is installed on your platform.

Which assistants does it track?

LightSite lists ChatGPT, Gemini, Claude, Perplexity and Google AI Overviews.

Is the Skills API used by ChatGPT or Claude?

LightSite records calls to it. Major assistants do not document calling third-party skills endpoints on ordinary websites, so check the logs for real usage.

Is it worth it for a Shopify store?

At $3,588 a year, it makes sense if you will use the tracking, bot analytics and execution help. A small store needing only mention tracking can pay less elsewhere.