Per our own AI-Overview audit of 1,026 buyer queries (June 2026), 879 returned an AI Overview and 233 of those cited FORKOFF.
FORKOFF · Tools
Enter a page, your brand and your category. The suite scores the page's AEO signals, checks whether AI crawlers may read your site, runs Lighthouse, and asks ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews whether they name you. One form, four reports, free.
<iframe
src="https://forkoff.xyz/tools/aeo-checker/embed"
width="100%"
height="1200"
frameborder="0"
loading="lazy"
title="AEO Checker by FORKOFF"
></iframe>Paste into a Custom HTML block.
Use /embed and paste the iframe src URL.
Drop into an Embed element. Lazy-loaded by default.
Three checks read your site and say whether answer engines can cite it. The fourth asks the engines and says whether they do. You need both: a page can be perfectly marked up and still never be named, and a brand can be named while its own site shuts the crawlers out.
Check 01 · Seconds
Can an answer engine lift a clean answer from this page?
Static HTML of one page, six signals: schema density 25, FAQ structure 20, definitional opener 20, content-signal headers 15, meta and canonical 10, AEO-extra 10.
Check 02 · Seconds
Are AI crawlers allowed in, and can they find what to read?
robots.txt access for seven AI bots (GPTBot, OAI-SearchBot, ChatGPT-User, Claude-Web, anthropic-ai, PerplexityBot, Google-Extended) 35, llms.txt 15, llms-full.txt 10, well-known manifests 20, sitemap 10, schema graph 10.
Check 03 · Under a minute
Is the page fast and clean enough to be fetched and trusted?
A mobile Lighthouse run: performance, accessibility, best practices and SEO, averaged. Weighted 20% in the composite beside page and agent readiness at 40% each.
Check 04 · One to three minutes
Do the engines name you when buyers ask?
Eight questions built from your brand, category and competitors, asked of ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews, scored 0-100 with a discovery gap.
The page-readiness check fetches one URL and scores six dimensions read straight from the static HTML: schema density, FAQ structure, definitional opener, content-signal headers, meta and canonical, and AEO-extra signals (llms.txt, JSON-LD alternate link, HTTPS). It answers whether the page's markup is structured the way an AI crawler expects; the citation check answers whether ChatGPT names the brand. The single highest-leverage fix for most sites is structured data, which our guide to schema markup for AEO covers signal by signal. If you are still deciding which surface to optimize first, the AEO vs GEO guide draws the extraction-versus-generative line, and our discovery gap research shows why brands cited for their own name still go missing on category questions.
The score sums six weighted dimensions read from the page's static markup: schema density (25 pts), FAQ structure (20), definitional opener (20), content-signal headers (15), meta and canonical (10), and AEO-extra signals (10). This check calls no LLM. It returns a 0-100 score, an A-F grade, and the finding behind every point lost.
4 checks
Coverage
5 AI surfaces
Asked live
3 points per JSON-LD block on the page (capped at 10), plus 2 points for every AEO-relevant schema type detected (FAQPage, Article, Organization, SoftwareApplication, Product, HowTo, WebPage, Service, LocalBusiness, VideoObject), capped at 15. FAQPage schema is the single strongest signal here.
Counts FAQPage schema Q&A pairs (up to 14 points) plus raw <details>/<dt> HTML patterns (up to 4 points), with a 2-point bonus at five or more total FAQ signals. Schema alone, HTML alone, or both all count.
Scans the first 200 characters of visible body text for a '{Brand} is {noun} for {ICP}' sentence. A full match scores 20, a partial 'X is...' without an ICP clause scores 10, no match scores 0. The cheapest fix on this checker.
Rewards exactly one H1 (7 points), two or more H2s (5 points), and any H3 (3 points). A hidden H1 (opacity-0 in the markup) is penalized 5 points, a pattern AI crawlers read as a manipulation attempt.
Title length (30-65 chars), meta description length (130-165 chars), and an HTTPS canonical tag score up to 10 points. A further 10 points come from a JSON-LD alternate link, an llms.txt / llms-full.txt reference, and HTTPS on the page itself.
Inputs · what the checker needs
One page URL. It drives page readiness, agent readiness and Lighthouse.
Your brand and category. They drive the eight questions the AI engines are asked.
Optional: up to three competitors.
Outputs · what the checker returns
Four scores and a composite. Page readiness, agent readiness and AI citations each score 0-100 with an A-F grade; the composite blends page, agent and Lighthouse 40, 40 and 20.
A full report per check. Every dimension shows its score against its maximum and the finding behind it; the citation report shows each question, each engine's answer and the competitors named.
A short fix list across the structure checks. The composite tab names the top fix from page readiness, agent readiness and Lighthouse, one each.
An AEO checker tells you whether answer engines like ChatGPT and Perplexity can read, trust and cite your site. This one runs four checks from one form. Page readiness scores the static HTML of one page on six signals. Agent readiness checks whether AI crawlers are allowed in and can find your llms.txt, manifests and sitemap. A mobile Lighthouse run covers speed and basic SEO. AI citations asks five AI surfaces eight buyer questions about your brand and records whether you are named. The first three say whether you are citable; the last says whether you are cited today.
Page-readiness patterns the suite sees often. Single audits diagnose; quarterly cadences fix.
Scenario 01
AWell-structured comparison page
A /compare/ page with FAQPage schema, a clear one-sentence definitional opener, and a clean H1-H2-H3 outline.
Composite score
88/100
Recommendation
Structurally sound. Hold this pattern on every new comparison page shipped.
Scenario 02
CLong-form blog post, thin structured data
A blog post with real prose depth and good headers, but no JSON-LD beyond the site-wide Organization schema, and no FAQPage block.
Composite score
54/100
Recommendation
Add FAQPage schema (up to 20 pts) and tighten the opening sentence into a direct definitional statement (up to 20 pts).
Scenario 03
BService page with rich schema, no FAQ
A /services/ page with Service and Organization JSON-LD, strong header structure, but zero FAQ content.
Composite score
71/100
Recommendation
Add an FAQ section with FAQPage schema, the single highest-leverage gap on this page type.
Scenario 04
FThin landing page, no structured data
A minimal launch page with hero copy and a CTA, no JSON-LD, no FAQ, no definitional sentence in the first 200 characters.
Composite score
12/100
Recommendation
Foundational gaps across every dimension. Add basic Organization or Service schema, an FAQ block, and a direct opening sentence before anything else.
Citable vs cited
Fixed rules. Live answers.
Every structure point traces to a fixed, published rule, so those scores only move when your site does. The citation score comes from the engines themselves and moves when they do. Fix the lowest structure score first, then rerun and watch whether the citations follow.
A rank tracker watches Google. A schema validator checks syntax. Asking ChatGPT by hand gives one answer to one question. The suite reads the site and asks five AI surfaces the same eight questions, with a finding behind every point.
← scroll horizontally to see more →
| Feature | FORKOFF AEO suiteFour checks from one form | Classic SEO rank trackersAhrefs · Semrush · Moz | Schema validatorsGoogle Rich Results Test and similar | Asking ChatGPT by handManual spot-check |
|---|---|---|---|---|
| Scores structured data (JSON-LD types and density) | ||||
| Scores FAQ structure and a definitional opener | partial | |||
| Checks whether AI crawlers are allowed in (robots.txt, llms.txt) | partial | |||
| Measures whether AI engines name the brand today | partial | |||
| Covers five AI surfaces with the same question set | ||||
| Repeatable scoring with a finding behind every point | partial | |||
| Best for tracking organic Google rank over time |
One audit tells you where the page stands today. These four habits decide whether the next audit reads better.
Before the first run, write down the eight questions a buyer would actually type and the answer you would want an engine to give. A question like “best clipping agency for a SaaS launch” is a real buyer question; “what is FORKOFF” is a vanity check that every brand passes. Keep the list fixed for a quarter so runs can be compared.
Then decide what a win is for each question. Being named in the answer is one bar; being named with the right category, the right price band and a link is a higher one. The citation score only means something when the team agreed on that bar before the numbers came back, not after.
Answer engines change their sources on their own schedule, so a single run is a snapshot, not a trend. Run the full suite once a month on the same pages and the same questions, and keep the exports side by side. A move of a few points between two runs is noise; the same move three months in a row is a direction.
While a fix is shipping, rerun only the structure checks on the changed page every week. Those scores trace to fixed rules, so they move the moment the page does and confirm the fix landed. The citation score follows later, often a few weeks after the engines recrawl, which is why the report keeps the two numbers apart.
Read every answer that names you, not only the count. Engines repeat old pricing, retired product names, a competitor's feature attributed to you, or a founder who left two years ago. Each of those is a sentence a buyer reads before they ever reach your site, and none of them shows up in a rank tracker.
When an answer is wrong, trace it to the page it came from. Usually it is an old blog post, a stale directory listing or a third-party review that still ranks. Fix the source you own, ask for the corrections you do not, and add the correct fact as a plain, quotable sentence on the page you want cited, with a date next to it.
The findings split cleanly between two teams. Schema, crawl access, the llms.txt file and page speed belong to whoever ships the site; they are one-time fixes per template and they lift every page on it. Definitional openers, FAQ answers, comparison tables and quotable figures belong to the content team; they are per-page work and they are where most of the citation gain comes from.
Hand each team its own list, sorted by the lowest score first, and agree on one owner for the monthly run. The pattern that stalls programs is a report that goes to everyone and is owned by no one: the web team waits for content, content waits for a template change, and the next audit reads exactly like the last.
For an agency or a team running several sites, keep one sheet per domain with the same eight questions per category. It turns a stack of one-off audits into a portfolio view, and it shows quickly which site needs the template fix and which needs the writing.
Month one fixes the template: schema on every page type, crawl access for the AI crawlers you want, an llms.txt that lists the pages you would like quoted. The structure scores jump within days because they read the page directly. The citation score usually does not move yet, and that is expected.
Month two is writing. Each money page gets a two-sentence definitional opener, a short FAQ written from the eight questions, and one figure a reader could quote with its date. Comparison pages get a real table instead of a paragraph of adjectives.
Month three is where the citation score starts to follow, first on the engines that browse live and later on the ones that lean on their training data. If the structure is green and citations have not moved after a full quarter, the gap is off the site: reviews, directories and the third-party pages the engines trust more than yours. That is the point to shift effort from the page to earned mentions: get listed where the engines already look, answer the questions people ask on Reddit and in forums, and keep the facts on those pages matched to your own.
If your decision lands in the left column, the checker is the right surface. If it lands in the right, use a different tool.
Built for this
Not the right fit
Four things from one form. Page readiness scores one page's static HTML on six signals: schema density (25 points), FAQ structure (20), a definitional opener (20), content-signal headers (15), meta and canonical (10) and AEO-extra signals such as llms.txt (10). Agent readiness checks robots.txt for seven AI crawlers, llms.txt, llms-full.txt, well-known manifests, the sitemap and the schema graph. A mobile Lighthouse run covers speed and basic SEO. AI citations asks ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews eight buyer questions and records whether your brand is named.
A tool that measures whether AI search engines name your brand when buyers ask about your category. The citation check here is one: it builds eight questions from your brand, category and competitors (best {category}, {brand} alternatives, {brand} pricing, {brand} vs a competitor and others), asks five AI surfaces, and scores how often you are named (weighted by engine), with a discovery gap that separates questions naming you from category questions where you are missing.
It is a 0-100 read of how often the five surfaces name your brand across the eight questions: a citation rate per surface, weighted ChatGPT 30, Claude 25, Gemini 20, Perplexity 15 and Google AI Overviews 10. Questions with your brand in them almost always cite you; the category questions are where the gap shows. A low score with strong structure scores means the engines can read you but are not pulling you into their sources: the fix is directories, comparison pages and third-party mentions, not more markup.
A GEO (generative engine optimization) audit checks whether AI crawlers can reach and understand a site. The agent readiness check here is that audit: robots.txt access for GPTBot, OAI-SearchBot, ChatGPT-User, Claude-Web, anthropic-ai, PerplexityBot and Google-Extended (35 points), llms.txt (15), llms-full.txt (10), well-known manifests (20), the sitemap (10) and the schema graph (10).
Composite = 0.4 x page readiness + 0.4 x agent readiness + 0.2 x Lighthouse, mapped to an A-F grade. The AI citation score is reported beside it rather than inside it, because it measures a different thing: whether you are cited, not whether you are citable.
The three structure checks usually return in under a minute. The citation check sends eight questions to five AI surfaces and takes one to three minutes. Scores appear as each check lands, so you can read the first reports while the citations finish.
A validator checks that your JSON-LD is syntactically valid. Page readiness scores whether the right types are present for answer engines, whether FAQ content exists, whether the page opens with a sentence an engine can lift, and whether headers and meta support extraction. Google retired the FAQ rich result in May 2026, but the FAQ signal still matters to answer engines that read the markup.
A sentence in the first 200 characters of body text that says what the brand is and who it is for, in a pattern like '{Brand} is {noun} for {ICP}'. Answer engines lift that kind of sentence directly when someone asks 'what is X'. A page that buries its definition three paragraphs down loses these points even when the content is strong.
Usually a wildcard Disallow rule, a security plugin or a CDN bot setting written before AI crawlers existed. The agent readiness report lists each of the seven bots as allowed or blocked, so you can see exactly which rule shuts which engine out.
The structure scores only change when your site changes, because they use fixed rules. The citation score can move between runs, because the AI engines change their answers. That is why a single run is a snapshot and citation share is worth tracking over time.
Not in one run. The structure checks grade the page you enter, so run each key page separately and compare the reports. The citation check is about the brand, so it does not change with the page.
Yes. The full run is free with no signup; an email unlocks the full reports and sends you a copy. A managed FORKOFF answer engine optimization engagement adds weekly citation tracking and does the fixes; talk to a strategist if you want that.
They are now the four tabs of this suite. Each old address redirects here, and every check they ran still runs, from one form instead of three.
Generate the Article, Dataset, Organization and FAQPage markup that the page-readiness fixes call for.
Where answer engine optimization ends and generative engine optimization begins, and which to fix first.
Why brands cited for their own name still go missing when buyers ask category questions.
Multi-LLM AEO scorecard · weekly cadence · sub-30% to 65% inside two quarters.
See how it works
Google is serving rank trackers poisoned SERPs. What one tracker measured, how to tell a fake drop from a real one, and how to report SEO and AEO you can trust.
Read
GEO is not replacing SEO. Here is what genuinely changed, what was only renamed, and measured crawler data on whether the two visibilities differ.
Read
AI SEO agency vs AI SEO software, decided with real Reddit AMA retainer figures, published tool pricing, and a stage-by-stage framework.
Read
