Category: Performance & Security

Performance guides for site owners: how many PHP workers your site needs, and how to get your website cited by ChatGPT and AI search.

  • PHP Workers Explained: How Many Does Your Site Need?

    PHP Workers Explained: How Many Does Your Site Need?

    A PHP worker is a server process that builds one uncached WordPress page at a time, so your worker count is how many uncached requests your site can handle at once. As of October 2026, Kinsta’s entry plans include 2 (it calls them PHP threads), Hostinger lists 40 to 100 by plan, Cloudways lists 100 per server on its Autonomous plans, and WP Engine and Liquid Web don’t publish a base count.

    This guide is based on each host’s published plans, documentation and terms, checked in October 2026; it does not include our own tests.

    What a PHP worker actually does

    When a request hits your site, one of two things happens. If the page is in cache, the server hands back a pre-built copy, with no worker involved. If it isn’t cached (a logged-in user, a cart, a checkout, a search, a REST or admin-ajax call), a worker picks it up, loads WordPress, runs your theme and every active plugin, queries the database, assembles the HTML, and only then frees up for the next request. Kinsta’s PHP docs put it plainly: each thread “processes one request at a time,” and “cached content doesn’t require PHP threads.”

    Think supermarket checkout lanes. Cached pages are the grab-and-go shelf by the door; everything else goes through a lane, and your worker count is how many lanes are open. With two workers, the third uncached request waits in line, no matter how powerful the rest of the server is.

    How many PHP workers do you need?

    Start with arithmetic. A worker that spends 1 second on an uncached request finishes at most one request per second; at 500 milliseconds, two. So at your busiest moment you need about your uncached requests per second multiplied by your average PHP response time in seconds, plus headroom. The real question is how much of your traffic bypasses the cache.

    • Brochure site or blog with full-page caching. Most pageviews come from cache. Cloudways’ PHP workers guide suggests 2 to 4 workers for a static site.
    • Busy blog with search, comments and logged-in authors. More requests skip the cache, so run the multiplication on your busiest hour.
    • WooCommerce, membership or LMS site. Kinsta’s pricing page recommends 4 or more PHP threads, and Cloudways calls 4 “a good starting point.” Cart, checkout and account pages bypass cache by design.
    • Busy store at sale time. Work from your peak: 6 uncached requests a second at 0.8 seconds each keep about 5 workers busy (6 × 0.8 = 4.8), before headroom.

    The multiplication that trips people up: worker capacity and code speed are the same lever. Four workers at a 1-second average response clear about 4 uncached requests per second, or 14,400 an hour. Let a slow plugin drag the average to 3 seconds and those same four workers clear about 1.3 per second. Your capacity just fell by two-thirds, and you didn’t change plans.

    PHP worker limits by host

    Here is what each host publishes on its own pages as of October 2026.

    Workers on entry planHow you get moreWhat happens at the limit
    Kinsta2 on Single 35k ($35 a month); 4 on Single 65k ($50)Bigger plan, or the prorated PHP performance add-on ($10 per 500MB a month)Requests wait; Kinsta warns of timeouts and 503 or 504 errors, and MyKinsta charts how often the limit is hit
    WP EngineNot published ($35 a month, or $28 for the first year billed annually)Not published; its 504 guide says to contact your account managerQueued requests are evicted with a 504; requests past 60 seconds are killed with a 502
    Liquid WebManaged WordPress: not published. VPS: whatever you configureElite autoscales up to 40 on the largest plans (first 24 hours a month included); on a VPS you tune PHP-FPMManaged WordPress: not published. VPS: CPU and RAM set the ceiling
    CloudwaysFlexible: not published. Autonomous (from $99 a month): 100 per serverFlexible: a bigger server. Autonomous: autoscaling adds serversNot published
    Hostinger40 on Premium ($2.99 a month on a 48-month prepay, renewing at $10.99)Unlimited (60) or Cloud Startup (100)Not published

    Notice the split. Kinsta and Hostinger publish a fixed count per plan (see our Hostinger review), and Kinsta lets you trade memory per thread for more threads. WP Engine and Liquid Web’s managed platform publish no base count, though Liquid Web‘s Elite tier adds workers during a surge. On a Liquid Web VPS or a Cloudways Flexible server you buy CPU and RAM, and Cloudways’ guide ties useful worker counts to CPU cores (a 2023 Cloudways post claims “up to 6000 PHP workers per site”; its pricing page gives no figure). That split is much of the argument in our Liquid Web vs Kinsta comparison.

    Why too few workers causes 504 errors

    The failure mode is a queue, not a crash. When every worker is busy, new uncached requests wait. Kinsta’s thread guidance for stores says that once requests outnumber threads, “the queue starts to reject new requests which could result in 502/504 errors.” WP Engine’s 504 guide is specific: when the workers and backlog are full, the oldest waiting requests are evicted with a 504, and requests still running after 60 seconds are cut off with a 502. On a plain nginx server, the default fastcgi_read_timeout is 60 seconds, so a request that waits longer for PHP ends in a 504.

    The signature: errors arrive in bursts under load, CPU can look normal, and dynamic pages fail first, so checkout breaks while the cached homepage loads. Common triggers are a sale or email blast landing at once, bots crawling uncacheable search and filter URLs, wp-cron pile-ups, and slow external calls, like a payment gateway, holding workers open. Confirm it in your host’s dashboard (MyKinsta has a PHP thread limit chart) or look for “upstream timed out” in the error log.

    How to need fewer workers

    Before you pay for more workers, make the ones you have finish faster. Halve your average response time and the same workers handle twice the uncached requests.

    • Get full-page caching right. Audit what’s excluded. A rule that skips every URL with a query string can make all your UTM-tagged campaign traffic uncacheable.
    • Add object caching. Kinsta’s PHP performance docs say Redis can “reduce the need for PHP threads.” Liquid Web includes Object Cache Pro on every managed WordPress plan, Cloudways includes it free on 4GB servers and up, and Kinsta charges $100 per site a month for Redis.
    • Hunt slow plugins and queries. Query Monitor or your host’s APM tool will name the offender. One slow plugin taxes every uncached request.
    • Move wp-cron to a real cron job. WordPress checks scheduled tasks on every page load, per its developer docs, which show how to hand that to the system scheduler.
    • Rate-limit the bots. Crawlers love faceted search and add-to-cart URLs, which can’t be cached and earn nothing. WP Engine already queues known bots behind real visitors.

    More workers, or a different host?

    On a host with fixed counts, price the next plan before you stack add-ons. At Kinsta, moving from Single 35k (2 threads) to Single 65k (4 threads) costs $15 a month more. The PHP performance add-on is prorated, so it suits a sale weekend; if it becomes permanent, compare it with the plan above. A resource-based server drops the fixed count but makes capacity your job. That trade-off, managed guardrails versus raw headroom, is the honest version of the is managed WordPress hosting worth it question, and for stores our best WooCommerce hosting for small stores guide covers where the line falls.

    For Kinsta against WP Engine, see Kinsta vs WP Engine; workers are only one line in that fight.

    Frequently asked questions

    What are PHP workers in WordPress?

    PHP workers are the server processes that build uncached WordPress pages. Each handles one request at a time: loading WordPress, running plugins, querying the database and returning HTML. Your worker count is the number of uncached requests your site can process at once.

    How many PHP workers do I need for WooCommerce?

    Kinsta recommends 4 or more PHP threads for ecommerce, membership and LMS sites, and Cloudways calls 4 a good starting point. Busy stores need more: multiply your peak uncached requests per second by your average response time in seconds, then add headroom.

    Do PHP workers matter if I use a caching plugin?

    Less, but they still matter. Caching removes most anonymous pageviews from the worker pool, which is why a well-cached blog can run on a 2-worker entry plan. Logged-in users, carts, checkouts, search and API calls still bypass the cache.

    Can too few PHP workers cause 504 errors?

    Yes. When all workers are busy, new requests queue, and Kinsta and WP Engine both document 502 and 504 errors when that queue overflows or waits too long. The tell: errors on uncached pages during traffic spikes while cached pages still load.

    Does adding more PHP workers make my site faster?

    Not directly. Kinsta’s docs say “more PHP threads do not automatically mean better performance.” Workers add concurrency: a page takes as long to build with 2 workers as with 10. More workers stop requests from queueing; if pages are slow with no traffic, fix the code and the caching.

    Want the guided version? Our free video courses are in production, starting with “Launch Your First Website with Claude.”

    Last checked: October 2026. Prices and plans change; check the vendor’s page before you buy.

  • How to Get Your Website Cited by ChatGPT and AI Search

    How to Get Your Website Cited by ChatGPT and AI Search

    To get your website cited by ChatGPT and AI search, let the search crawlers reach your pages (OAI-SearchBot for ChatGPT, PerplexityBot for Perplexity, Googlebot for Google’s AI Overviews and AI Mode), serve those pages quickly, and write content that answers the question clearly. As of October 2026, Google says there are “no additional requirements to appear in AI Overviews or AI Mode,” and an llms.txt file does nothing for Google Search.

    This guide is based on the AI providers’ own documentation and one large published study, checked in October 2026; it does not include our own tests.

    Getting cited by AI search overlaps heavily with classic SEO, and a lot of what’s published about it is untested theory. Here’s what the providers document, what the data shows, and where your effort is best spent.

    How AI search actually picks its citations

    Your site shows up in an AI answer through one of two doors. The first is training data: the model absorbed your content months ago. You can opt out of that (GPTBot, ClaudeBot and Google-Extended are the controls), but you can’t steer it. The second is retrieval, and that’s the one you can win. OpenAI says OAI-SearchBot “is used to surface websites in search results in ChatGPT’s search features.” Perplexity says PerplexityBot “is designed to surface and link websites in search results on Perplexity.” Google says AI Overviews and AI Mode retrieve pages from its Search index and may use “query fan-out,” a set of concurrent, related queries, to gather more supporting pages.

    That gives you three things to get right: the crawler has to be allowed in, your page has to be indexed and eligible (for Google, “indexed and eligible to be shown in Google Search with a snippet”), and your content has to answer the question clearly enough to be worth citing.

    One prerequisite before anything else: check robots.txt and your firewall. For citations, the bots that matter are the search ones: OAI-SearchBot, PerplexityBot, Claude-SearchBot and Googlebot. GPTBot, ClaudeBot and Google-Extended govern AI training, not search results: OpenAI says each of its settings “is independent of the others,” and Google says Google-Extended “does not impact a site’s inclusion in Google Search.” Google’s AI features guide lists “Ensuring that crawling is allowed in robots.txt, and by any CDN or hosting infrastructure” among the basics, and OpenAI and Perplexity both recommend allowing their published IP ranges, so check that a security plugin or WAF rule isn’t blocking them. In Google Search Console, the Search generative AI control under Settings should be set to include your site, which is the default.

    llms.txt: does it actually work?

    llms.txt is a proposal, first published by Jeremy Howard in September 2024 and revised in August 2026, “to standardise on using an /llms.txt file to provide information to help agents use a website”: a markdown file, placed at the site root or under a path, with brief background, guidance and links to your most useful content. It’s quick to create, and some AI-SEO advice treats it as essential.

    The evidence says it isn’t. Google’s AI optimization guide says you don’t need AI text files to appear in Google Search “as Google Search itself doesn’t use them,” and that maintaining one “will neither harm nor help your site’s visibility or rankings in Google Search.” For ChatGPT, SE Ranking’s November 2025 study of 129,000 domains found llms.txt “has shown negligible impact on ChatGPT citation likelihood.” The proposal notes that OpenAI, Anthropic and Gemini publish llms.txt files for their own developer docs; that’s publishing one, not a promise to read yours.

    Ship one anyway if you like: it’s harmless, and coding agents that read documentation may use it. But it’s dessert, not dinner. Here’s where the effort-to-impact ratio sits, based on the sources in this guide:

    TacticEffortWhat the sources say
    Allow AI search crawlers in robots.txt and your firewallMinutesPrerequisite: OpenAI says sites opted out of OAI-SearchBot “will not be shown in ChatGPT search answers”
    Links from other sitesOngoingNumber of referring domains topped SE Ranking’s list of ChatGPT citation factors
    Fast pagesDepends on your hostCorrelated with ChatGPT citations in SE Ranking’s study, with diminishing returns
    Clear, well-organized pagesOngoing disciplineGoogle recommends clear headings; question-based headings correlated with citations in SE Ranking’s data
    Structured dataLowNot required for Google’s AI features; pages with FAQ schema averaged fewer ChatGPT citations in SE Ranking’s data
    llms.txtMinutesIgnored by Google Search; negligible impact on ChatGPT citations per SE Ranking
    Third-party “AEO” or “GEO” toolsPaid subscriptionGoogle: “No third-party tool has access to our internal ranking or AI systems”

    Speed: the citation factor to budget for

    SE Ranking’s study of 216,524 pages on 129,000 domains found that pages with a first contentful paint under 0.4 seconds averaged 6.7 ChatGPT citations, against 2.1 for pages slower than 1.13 seconds; SE Ranking describes fast pages as “3 times more likely to be cited.” It’s a correlation, not a controlled test, and the returns diminish: pages in the 0.65 to 0.82 second range held steady at 4.2 citations, and referring domains and traffic ranked above speed among the factors it measured.

    Two documented facts explain why raw server speed matters. Google lists “reducing latency” as part of a good page experience, and Google’s web.dev guidance treats a time to first byte of 0.8 seconds or less as good. And not every AI crawler runs JavaScript: Vercel’s December 2024 analysis found that ChatGPT’s and Claude’s crawlers fetch JavaScript files but “don’t execute them,” so content that only appears after JavaScript runs may never reach them. Google, by contrast, says it can process JavaScript “as long as it isn’t blocked.”

    Caching changes the picture. A crawler that requests a page your cache isn’t holding gets your uncached origin performance: PHP execution, database queries, worker availability. If you’ve read our PHP workers guide, this is that same bottleneck wearing a new hat.

    This is where hosting stops being a commodity decision. Pages that load slowly at the origin are harder for any crawler to fetch, so server-level caching matters. Managed WordPress hosts document it: Kinsta serves sites through Cloudflare with edge caching, and WP Engine documents a Varnish page cache plus Cloudflare edge full-page caching, with entry plans at $35 a month on both. Our managed WordPress hosting comparison and the AI web hosting pillar cover the options. For AI visibility, treat origin speed as something you choose with your host, not something you tune into existence on a $4 plan.

    Write pages that answer the question

    Google says you don’t need to write in a special way for AI search, break content into tiny chunks, or rewrite pages just for AI systems. What it recommends is content organized “by paragraphs and sections, along with headings that provide a clear structure.” SE Ranking’s ChatGPT data points the same way: longer articles, Q&A sections and question-based headings correlated with more citations, and pages updated in the past three months averaged 6 citations against 3.6 for outdated ones. In practice:

    • Answer first. Put the direct answer at the top of the page and under each heading. Context and caveats come after.
    • Headings as questions. Phrase headings the way a person actually asks; question-based headings correlated with citations in SE Ranking’s study.
    • Lists and tables. Use them where the content is a list or a comparison. They’re easier for readers to scan.
    • Focused sections. One topic per section helps readers. Google says its systems can “understand the nuance of multiple topics on a page,” so this is for people first.
    • Visible dates and a named author. Show who wrote the page and when it was last updated, and keep it current: recently updated pages averaged nearly double the citations in SE Ranking’s data.

    None of this fights classic SEO. In Google’s words, optimizing for generative AI search “is optimizing for the search experience, and thus still SEO.”

    Structured data: useful, not required

    Structured data won’t rescue a slow, muddled page. Google says it “isn’t required for generative AI search, and there’s no special schema.org markup you need to add,” though it still helps with rich results in regular search. If you use it, make sure it “matches the visible text on the page.” Skip FAQPage markup as an AI tactic: Google no longer shows the FAQ rich result in Search, and in SE Ranking’s ChatGPT data, pages with FAQ schema averaged 3.6 citations against 4.2 without.

    On WordPress, Yoast and Rank Math both generate schema markup; validate it with Google’s Rich Results Test rather than trusting the plugin’s green light. And if you want AI search on your own site, WP Engine’s Smart Search AI adds AI-powered semantic search, and WP Engine says it now works on WordPress sites “regardless of hosting provider.” That’s a different job than external citations; setup is in our Smart Search guide.

    How to know it’s working

    Google now reports this directly. Search Console’s Generative AI performance report, rolled out to all websites on August 31, 2026, shows impressions from AI Overviews and AI Mode by page, country and device. For ChatGPT and Perplexity, measurement is manual. Three habits cover it:

    • Grep your access logs monthly for OAI-SearchBot, PerplexityBot, Claude-SearchBot and Googlebot to confirm the search crawlers are reaching your pages. GPTBot and ClaudeBot visits are training crawls.
    • Segment AI referrals in analytics. Filter for chatgpt.com and perplexity.ai referrers to see how many visitors arrive from AI answers.
    • Ask the money questions yourself. Once a month, put your most valuable queries to ChatGPT, Perplexity and Google and record who gets cited. It’s tedious, and it’s the most direct check available.

    Expect some lag. OpenAI says it can take about 24 hours for its systems to adjust after a robots.txt change, and Google says a change to the Search generative AI control generally takes a few days to apply. Beyond that, citations depend on crawling, indexing and ranking, which no one can schedule. This is a compounding channel, not a campaign.

    Frequently asked questions

    How do I get my website cited by ChatGPT?

    Allow OAI-SearchBot in robots.txt and in your firewall (OpenAI says sites opted out of it “will not be shown in ChatGPT search answers”), make sure pages load quickly and don’t depend on JavaScript for their main content, and put a clear answer at the top of each page. GPTBot is OpenAI’s training crawler; blocking it doesn’t remove you from ChatGPT search.

    Does llms.txt actually improve AI visibility?

    Not for Google: Google Search ignores llms.txt files, and maintaining one “will neither harm nor help” your visibility there. SE Ranking’s study of 129,000 domains found negligible impact on ChatGPT citations. It’s harmless to add; just spend your real effort on crawler access, speed and content first.

    How long does it take to show up in AI search results?

    There’s no fixed timeline. Google’s AI Overviews and AI Mode draw on pages in its Search index, so a page has to be crawled and indexed first, and OpenAI says robots.txt changes take about 24 hours to reach its systems. Getting absorbed into a model’s training data isn’t something you can schedule or plan around.

    Should I block AI crawlers instead of courting them?

    You can block training without leaving AI search. Disallowing GPTBot or ClaudeBot opts your content out of those companies’ model training, and Google-Extended covers Gemini training and grounding in the Gemini app, but none of them affects OAI-SearchBot, PerplexityBot or Google Search. Blocking the search crawlers removes you from those answers. If your business is selling the content itself, weigh that; for most site owners, AI answers are becoming a discovery channel.

    Last checked: October 2026. Prices and plans change; check the vendor’s page before you buy.

    Want the guided version? Our free video courses are in production, starting with “Launch Your First Website with Claude.”