A vertical-specific look at what Foxy does on content sites with deep archives: where it beats keyword search, where it does not help, and how to set it up on WordPress, Ghost, Substack-published, or a custom CMS.
Info sites and content publishers live or die on archive discoverability. A site that has been writing for five years has hundreds or thousands of articles, and most of the value is in the back catalogue. Readers come in through a single article from search or social, and the question is what happens after they finish reading. Most of them leave because they cannot find the next thing they want to read, even though that next thing exists on the site — it is just buried three categories deep.
Site search on info sites is notoriously underpowered. Most CMS-default search runs on simple keyword matching, ignores synonyms, ranks by date instead of relevance, and dumps a long list of titles on the reader who then has to skim each one to figure out which is the article they actually want. Power users learn to skip the site search entirely and use Google with a "site:" filter, which is a tacit admission that the site cannot find its own content.
The deeper problem is that readers do not search by title, they search by question. A reader asks "what did this writer say about remote work in 2024?" or "is there an article on this topic for beginners?" or "what is the difference between this approach and that approach?" Those questions cannot be answered by a keyword box. They need a system that understands the question, knows the archive, and can pull together the right two or three articles into a coherent recommendation.
Foxy reads your full article archive, your author pages, your topic indexes, and your about pages, and answers reader questions in plain language with the source still attached. The retrieval is hybrid: semantic search ranks articles by meaning so synonyms and paraphrases work, and full-text search catches specific titles, author names, and quoted phrases. A reader who asks "did you cover the remote-work shift in 2024?" gets a direct list of the three most relevant articles with one-sentence summaries, not a search results page.
Foxy understands site structure when you let it. Category pages, author pages, tag pages, and chronological archives all become navigable through chat. A reader can ask "show me everything by this author about that topic" and get a curated response that respects both filters. Same for "what did you publish last month on X" or "do you have a beginner guide to this thing". The chat surface effectively replaces the navigation menu for readers who would rather ask than browse.
Where the archive does not have a strong answer — the reader is asking about a topic the publication has not covered, or about a piece behind a paywall the visitor cannot access — Foxy says so honestly. It does not fabricate articles, does not pretend coverage exists where it does not, and offers to capture the visitor's interest as a content-gap signal the editorial team can review. That gap surface is the second most-loved feature among publishing customers, right behind the article search itself.
Three reader surfaces compound. Article search is the heaviest-lift one. A reader who just finished an article and wants more on the same topic gets a list of three to five next-reads pulled from semantic similarity, not just shared tags. The lift in pages-per-session is visible in analytics the first week. The lift in newsletter signups is often visible the second week, because readers who find the next article are also more likely to subscribe.
Related content is the second. Foxy can answer "what should I read after this?" mid-article, with recommendations that respect both topical similarity and reading level. For sites with educational content, that means a beginner reader gets pointed to other beginner pieces, and an advanced reader gets pointed to other advanced pieces — without anyone having to manually tag every article with a difficulty level.
Author and topic discovery is the third. Readers often discover an author from one piece and want to find everything else they have written. Foxy can answer "show me everything by this author" with the full list and a one-sentence summary of each piece. Same for topic discovery: a reader who liked one piece on a niche topic can ask "what else have you written about this?" and get a curated thread that no auto-generated related-articles widget could match for quality.
A fourth surface that pays back over time is newsletter and editorial-calendar feedback. The questions readers ask Foxy that the archive cannot answer are the highest-signal editorial brief most publishers will ever see. They are the topics your audience is hungry for and nobody on the team has covered yet. The unanswered-questions queue clusters those gaps by topic so the editorial team can scan a single weekly digest and pick the three or four that are worth commissioning. Some publications turn the digest into a public-facing "ask the editors" column. Others quietly use it as the input to next quarter's content plan. Either way, the loop tightens: readers ask, Foxy logs the gap, the team publishes the piece, Foxy answers it the next time the same question lands. The archive grows where the audience is paying attention.
Paywall enforcement is explicitly outside Foxy's lane. If your site has paid articles, Foxy can describe what is in them, link to them, and tell visitors how the subscription works, but it should not be summarising or extracting the paid content itself for free readers. FoxChat ships a per-URL access-tier setting so you can mark which URLs are gated and configure Foxy to respect that gating. The summary visitors see for paid pieces stays at the level of a teaser, never the substance.
Subscription support is the other category Foxy does not handle alone. Failed payments, plan changes, refund requests, cancellation flow, and account-recovery questions are operator territory. Foxy can answer the public policy questions — how much, how to cancel, when do you bill — but a specific reader asking about their specific charge needs an operator with access to the subscription platform. The handoff is clean, with the policy already shared and the reader's question already framed, so the operator picks up a triaged thread.
The third boundary worth naming is opinion and commentary that is not in the archive. Readers sometimes ask "what do you think about this news story?" or "do you have a take on this debate?" and Foxy will not invent an editorial position the masthead has not actually published. The honest answer is that the publication has not covered the topic, with an offer to capture the reader's interest so the editors can see the gap. That restraint is what keeps a publisher-facing chat from accidentally becoming a generative-AI rumour mill.
WordPress. The FoxChat script goes into header.php or footer.php, or via the script-tag injection of any decent WordPress plugin. The dedicated WordPress install guide walks through theme editing, Gutenberg block compatibility, and how to ensure the widget loads after your post content rather than during first paint. See the WordPress install guide.
Ghost, Substack-published, and custom CMSes. Ghost takes the script in code-injection settings. Substack-published sites (custom domains served through Substack's renderer) can inject the widget via the site customisation surface. Custom CMSes — Eleventy, Hugo, Astro, Sanity-rendered fronts, Strapi-backed Next.js, anything else — get the same script tag in the base template. The crawler respects your sitemap and the standard schema.org Article markup, so the archive is indexed without manual page-by-page upload. For sites with very deep archives (10,000+ articles), the initial index takes a few hours; incremental updates run continuously after that. See the full install guides.
The 14-day trial gives you a fully indexed archive within minutes of pointing the crawler at your sitemap. Foxy answers reader questions about your back catalogue from day one, and the unanswered-questions queue surfaces the content gaps your editorial calendar should probably address next. Start your trial, see the full feature list, or check the pricing page first.
No credit card. One script tag. Foxy is recommending the next read from your archive within an hour.
Start your free trial See the demo