AI Search Optimizer
Track citations across ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews + Grok — the new SEO.
What it is
GEO (Generative Engine Optimization) is the new SEO — AI chatbots cite specific sites when answering questions, and you want yours in the answer. Outerank's GEO module tracks your brand across SIX consumer AI surfaces: ChatGPT, Perplexity, Google Gemini (grounded), frontier AI the AI, Google AI Overviews (via a live SERP integration), and xAI Grok (live search). Five tabs. Overview: 7-point readiness scan + score, full report with the AI's narrative fix list, citation-rate trend, and a Share-of-AI-Voice chart that stacks you against your top 3 competitors. Citations: multi-engine tracking with real APIs (engine-status strip shows which keys are wired), per-engine breakdown, mention-vs-cited distinction, competitor leaderboard on the SAME prompts, custom prompts or 'Discover' for 10 AI-generated suggestions, and click-to-fix recommendations. Insights (NEW): the same citation data sliced four different ways — By URL (rank your OWN pages by how many AI engines cite them), Recently Cited (a strength-ranked feed of your AI-search wins with cross-engine confidence badges), Agreement (per-prompt confidence score showing how many engines corroborate a citation), and URL Discovery (sweep 25 / 50 / 100 prompts and surface pages winning in AI search you didn't know were winning). Honest DEMO mode when no engines connected — banner clearly tells you the citations are the AI estimates, billing is waived. Schemas + AEO: FAQ / HowTo / Article / Product JSON-LD generators (FAQ auto-suggests Q&As from any URL), plus an AEO content rewriter. Tools: llms.txt + entity JSON-LD + AI snippet preview + OG scan + search-action schema + llms-full.txt — each with a Deploy button for connected CMS (WordPress/Shopify/Webflow/Ghost/HubSpot).
Why it matters
Searches are migrating to AI chatbots. The sites that win citations get traffic AND brand authority; the sites that don't become invisible in the answer box. Most SEO tools haven't caught up — they sell 2019-era backlink reports. Outerank GEO is built for the AI era from the ground up: real multi-engine citation tracking, competitor benchmarking, schema generators, and an AEO rewriter that physically restructures content to be quotable. This is the moat.
How to use it (step-by-step)
Follow these in order. Each step is 30 seconds to 2 minutes.
Step 1
Read the honest-mode banner first (Overview tab)
If you see an amber 'DEMO MODE · No AI engines connected' banner at the top, every citation number below is the AI estimating, not real ChatGPT/Perplexity/etc responses. Demo mode is FREE — your credits aren't charged — but never report demo results to clients as real data. Connect 1+ engine in Settings → AI Engines to flip the module to real-data mode (credits start applying).
Step 2
Run a GEO scan (free, Overview tab)
Checks 7 AI-readiness signals (schema, llms.txt, E-E-A-T) and gives you a 0-100 readiness score. The the AI narrative recommendation is collapsed by default — click to expand for the fix plan.
Step 3
Check the Cockpit's 'What to do today' card
The big emerald card on the right of the Cockpit picks the highest-leverage action for your current state: connect an engine · pick a project · run first check · open AEO Rewriter (low rate) · build FAQ schema (engine gap). One click, you're on the right tool.
Step 4
Check AI citations on YOUR questions (Citations tab)
Switch to 'Custom' mode, paste up to 10 real customer questions OR click 'Discover 10 ideas' (1cr lavender button) and pick. We run them on all 6 AI engines (ChatGPT · Perplexity · Gemini · the AI · Google AI Overviews · Grok) for engines that are connected, score you AND your tracked competitors, and surface a Profound-grade results table with engine pills per row.
Step 5
Read the Profound-grade citation table
One row per prompt. Engine pills show which of the 6 engines cited you (emerald) vs mentioned (amber) vs absent (grey). Competitor chips on the right show which rivals got cited on the SAME prompt. Click 'Rewrite' on any absent row to open AEO Rewriter pre-positioned.
Step 6
Run Deeper Analysis (Moats panel)
Below the citation table, three buttons unlock the moats no competitor ships: 'Analyze sentiment' (1cr · per-prompt positive/neutral/negative + verbatim phrase ChatGPT used to describe your brand) · 'Cluster topics' (1cr · groups your prompts into 3-7 semantic buckets with per-bucket rate · 'you win on comparisons 80% but lose on tutorials 0%') · 'Quality score' (FREE · how prominently you're cited — position in answer + word distance + linked-vs-mentioned).
Step 7
Check the Brand Authority Graph (Overview)
The 8 platforms AI engines mine when deciding who to cite (Wikipedia · LinkedIn · Crunchbase · G2 · Trustpilot · Capterra · Product Hunt · Glassdoor). Hit 'Check authority graph' (3cr · cached 7 days). Each platform shows Listed (open profile) or Claim (signup URL). Weighted score 0-100 reflects how well-known you are to AI.
Step 8
Generate schemas for your top pages (Schemas + AEO tab)
FAQ schema is the single biggest GEO booster — paste your page URL, Outerank suggests 5-7 Q&A pairs (1cr · refund-on-fail), you edit, you get a paste-and-go JSON-LD snippet. Also HowTo / Article / Product generators (FREE, pure code).
Step 9
Rewrite weak pages with the AEO rewriter (Schemas + AEO tab)
For each prompt where you weren't cited, find the page that targets that intent, paste a paragraph, hit 'Rewrite for AI' (2cr · authorship-leak guarded). the AI restructures it: clear answer first, 3 supporting bullets, citable stat — the shape AI engines extract verbatim. Refund-on-fail if the guardrail blocks the output.
Step 10
Deploy your artifacts (Tools tab)
Generate llms.txt + entity JSON-LD (both FREE), then hit 'Deploy' — we detect your connected CMS (WordPress/Shopify/Webflow/Ghost/HubSpot) and give you the exact paste-in spot. Also AI Snippet Preview (1-prompt drill-down: 'describe my brand') and OG scan + search-action schema.
Step 11
Schedule the auto-check (weekly cron @minute=35)
Set a per-project cadence (weekly · biweekly · monthly · off) via POST /schedule. The geo_schedule_dispatch cron picks up due schedules every hour at minute=35 and runs check_citations() automatically. Trend line + Brain Forge signals (geo_low_visibility · geo_engine_gap) update automatically.
Step 12
Open the Insights tab — By URL view
The By URL section groups the latest citation result by YOUR domain's pages. Each row shows which prompts cite that page and across which engines, with a confidence badge (none/low/medium/high/max) based on how many engines out of the ones you ran corroborate the citation. Rank pages by AI-search authority, not just guess.
Step 13
Read the Recently Cited feed (Insights)
Flat strength-ranked list of every (prompt × engine) hit from the latest run. Cited beats mentioned, and cross-engine cited items float to the top. Each row shows the exact answer snippet the engine returned + the URL(s) of yours that got the citation — the raw material for a Receipts Wall on your marketing site.
Step 14
Check the Cross-engine Agreement score (Insights)
Per-prompt table: how many of the 6 engines you ran actually cited you for THIS prompt. A prompt cited by 4/6 engines is a much stronger authority signal than one cited by 1/6, even if the 1/6 engine is ChatGPT. Confidence badges (low/medium/high/max) turn the raw ratio into an at-a-glance read.
Step 15
Run URL Discovery — wider net (Insights, Pro+)
The regular Citation Tracker runs 5–10 prompts; URL Discovery runs 25 (Pro), 50 (Studio), or 100 (Agency) and groups the results by which of YOUR URLs get cited. Finds citations you didn't know existed. Charged as normal citation-check units (prompts × engines) with partial refund on failed pairs. Same DEMO-mode honesty — sim-only runs are FREE. TWO-STEP CONFIRMATION: the Run button always asks 'About to spend up to Xcr — Confirm & run?' so a misclick or double-click never drains credits. Every prompt that actually ran is listed after the run (copy or reuse for next time — no regeneration cost). Optional 'Targeting hint' field steers the auto-generator for this ONE run without touching your project description.
Step 16
Read the engine-icon colour code (Citations table)
Every row shows the 6 engines as small icons in the 'Cited by' column, colour-coded by status: GREYED = ABSENT (that engine's answer neither mentioned nor linked your domain) · AMBER/orange = MENTIONED (the AI said your brand name in the answer text but didn't link) · EMERALD = CITED (your domain is in the AI's source list — the strong signal). Click any icon to swap the expanded answer to that engine's response and see exactly why it cited a competitor instead of you.
Step 17
Browse every past run (Insights → Run history)
Every citation check (Citations tab AND URL Discovery) is now saved in full and browsable at the top of the Insights tab. Click any past run to time-travel — By URL / Recently Cited / Agreement all update to show that run's data. The bar chart on Overview keeps every run's rate; the Insights tab keeps every run's per-prompt detail. Older runs from before 2026-09-09 show as counts-only (no payload was saved back then).
Step 18
Re-check + watch the trend climb
The Overview shows your citation-rate trend as bars (one per check, real-data vs demo distinguished). We auto-alert you (in-app + email) if your rate drops more than 20 points run-over-run.
Are AI engines actually citing you?
Outerank runs your priority prompts through ChatGPT, Perplexity, Gemini, and Google AI Mode — then checks if your site is cited as a source. The citation rate is the AI-search ranking signal you can't see anywhere else.
- Multi-engine: ChatGPT · Perplexity · Gemini · AI Mode
- Real prompts, real answers, real citations
- 10 credits per prompt, results saved
- Compare prompts over time as engines evolve
Coverage by engine
Schema + canonical + entity links — graded.
AI engines parse structured data more aggressively than Google's classic crawler. Outerank grades your JSON-LD coverage, canonical correctness, hreflang sanity, entity-link density — the structural signals AI engines actually read.
- JSON-LD per page (Article / FAQPage / Product / Org)
- Canonical + hreflang validation
- Entity-link density (Wikipedia-style links)
- FAQ block presence (AI engines love these)
Structural grade
Share of AI voice — vs your top 3 competitors.
On the same prompt set, who else is being cited? Outerank ranks the top 3 competitors by AI-citation rate so you see exactly where the AI-search game is being won. The leaderboard is the gap-analysis you can't get elsewhere.
- Side-by-side citation rates on identical prompts
- Competitor entity-link density compared
- Engine-by-engine leaderboard
- Brain flags significant share-of-voice swings
Share of AI voice
Brain prescribes the exact AI-search fix.
Click 'Analyze' and Brain reads your citation report + structural audit, then writes a priority-ordered fix list. "Add Article schema to /blog/* (high ROI)", "Add a FAQ block to /products/foo" — specific, ranked.
- Priority-ordered fix list, not a generic checklist
- References your real URLs
- Estimated lift % per fix
- Saved analyses are free to re-open
Brain · GEO fix list
P1 · Add FAQPage schema to top 12 pillar pages — +18% citation rate.
P2 · Wire entity links (Wikipedia) on /glossary — +9% on Perplexity.
P3 · Fix 3 hreflang errors blocking AI Mode parsing.
Why Outerank
What classical SEO tools don't measure yet.
AI-search isn't "future SEO" — it's already a third of informational queries. Outerank measures it natively.
Real multi-engine citation rate
ChatGPT + Perplexity + Gemini + AI Mode, all four checked. Most tools query one engine and call it done.
Share of AI voice leaderboard
Side-by-side vs your top 3 competitors on identical prompts. The gap analysis nobody else surfaces.
Structural-signal grader
JSON-LD coverage, canonical, hreflang, entity links — the structural signals AI engines actually parse.
Brain writes the priority list
Not a 50-item checklist. A priority-ordered fix list with estimated lift % per item, referencing your real URLs.
Trend over time
Citation rate is a moving target. Outerank tracks it so you can see when an engine learned about your content.
Free AI Search Checker
Anyone can run a 1-prompt check before signing up. The on-ramp for the entire moat.
What costs credits
No surprises. Browsing and viewing saved results is always free.
Pro tips
- →Connect Perplexity Sonar first — it's the cheapest real engine ($0.005/query) and gives the strongest citation signal. One key flips the whole module to real-data mode for trial customers.
- →Custom prompts beat auto-generated. Paste 5 questions YOUR real buyers ask — that's the AI search ground truth.
- →Run the Brand Authority Graph before anything else — if you're missing from Wikipedia + Crunchbase + G2 + LinkedIn, no amount of FAQ schema will fix you. AI engines need authoritative sources to cite from.
- →Use sentiment analysis after a real-mode citation check — knowing ChatGPT calls you 'expensive but powerful' vs 'budget option for startups' is gold for positioning.
- →FAQ schema is one of the strongest single GEO levers we track. In our observations, pages that add FAQPage JSON-LD tend to be cited noticeably more often once the page also answers the question directly. Add it to your homepage AND every key landing page.
- →Use the AEO rewriter on your top 10 blog posts that aren't getting cited. Rewriting just the intro paragraph (answer first, bullets after) often flips them into AI snippets.
- →Track competitors that beat you on the same prompts. Open their pages, study how they structured the answer, then rewrite yours sharper.
- →Run citations weekly once you have things shipping — the trend line is the proof your work is paying off, and the alert system catches drops before they hurt you.
- →Generate llms.txt + entity JSON-LD early — these are free, take 10 seconds, and most sites still don't have either.
- →Open Insights → By URL after every real-mode check. Any URL with a 'high' or 'max' confidence badge is a page you should market as 'cited by AI' on your homepage / launch page — screenshots + timestamps beat any 'trusted by X companies' claim.
- →Use URL Discovery once a month on 50–100 prompts — it uncovers the long-tail prompts you're already cited for, most of which you didn't know existed. Feed the winners into your ad copy.
Plain-English glossary
Every bit of jargon on this page, explained.
- GEO
- Generative Engine Optimization — making your site cited by AI chatbots (ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews, Grok), the way SEO targets Google's blue links.
- AEO
- Answer Engine Optimization — structuring content so AI engines can extract a clean, quotable answer. Lead with the claim, follow with evidence, end with attribution.
- Demo mode (honest mode)
- When 0 AI engine keys are connected, the citation tracker falls back to the AI estimating what each engine WOULD say. Amber banner makes it explicit, billing is waived — your credits are not charged in demo mode.
- Citation vs Mention
- Citation = your domain is in the engine's citation list (the strong signal). Mention = your brand name appears in the answer body without a link (still valuable for AI memory).
- Citation rate
- Of the prompts tested, the percentage where you got CITED in at least one engine. Higher = more AI visibility.
- Per-engine breakdown
- Citation rate split per engine — you might be cited 80% on Perplexity but 0% on Gemini. Tells you where to focus.
- Competitor leaderboard
- Same prompts, ranked by who got cited most. Shows exactly who's winning the AI search game in your niche.
- Share of AI Voice
- Your citation % vs your top competitors on the same prompts, charted over time.
- Brand Authority Graph
- Real live presence check across the 8 platforms AI engines mine when deciding who to cite (Wikipedia · LinkedIn · Crunchbase · G2 · Trustpilot · Capterra · Product Hunt · Glassdoor). Weighted score 0-100. 7-day cache. The 'do you exist to AI engines' diagnostic.
- Mention sentiment
- Per-cited prompt classification of the brand mention as positive / neutral / negative + extracted verbatim phrase. Knowing ChatGPT calls you 'expensive but powerful' is gold for positioning.
- Topic clusters
- Groups your tracked prompts into 3-7 semantic buckets with per-bucket citation rate. 'You win on tool comparisons (80%) but lose on tutorials (0%)' beats a single average that hides everything.
- Answer Quality Score
- Prominence scoring 0-100 per citation: position in answer (1st = 100, 5th = 40) + word distance to brand mention + linked-vs-mentioned weight. Deterministic, free.
- Cross-engine agreement (confidence)
- Of the engines you ran, how many actually cited you for this prompt. Expressed as a ratio (cited_engines / engines_run) and bucketed into confidence tiers: none (0), low (1/6), medium (2/6), high (3–4/6), max (5–6/6). Higher agreement = the AIs converge on your page as the answer — the strongest possible authority signal.
- URL Discovery
- A wider citation sweep — 25 / 50 / 100 prompts (plan-gated Pro / Studio / Agency) versus the standard 5–10. Groups results by which of YOUR URLs get cited so you find long-tail prompts you're already winning on but didn't know were relevant. Charged at normal citation-check rates with partial refund on failed pairs. Two-step confirmation guards the Run button; every prompt that ran is listed after with Copy + Reuse buttons; optional Targeting hint steers the auto-generator per run.
- Engine icon colours (Citations table)
- Greyed = ABSENT (that engine did not cite or mention you) · Amber = MENTIONED (brand name in the answer text, no link — still valuable for AI memory) · Emerald = CITED (your domain in the source list — the strong signal that drives traffic). Every icon is clickable — click it to read that engine's exact answer + its own citations for that prompt.
- Run history (Insights tab)
- Every citation check (Citations tab OR URL Discovery) is saved in full and browsable at the top of the Insights tab. Pick any past run to time-travel — By URL / Recently Cited / Agreement all update. Snapshots before 2026-09-09 show as counts-only.
- Targeting hint (URL Discovery)
- Optional one-line field that steers the auto-generated prompts for THIS run only (e.g. 'focus on GEO wedge; competitors are Ahrefs/Semrush; buyer is agency owner'). Cheaper than updating your project description just to test an angle. Ignored when you paste your own prompts.
- AI Snippet Preview
- Single-prompt diagnostic asking AI 'Describe my brand in 3 sentences.' Different from Citation Tracker (5-prompt batch) — this is your brand-identity pulse-check.
- llms.txt
- A plain-text file at your domain root (like robots.txt) telling AI crawlers which pages matter. Emerging standard, low effort, high signal.
- FAQ schema
- JSON-LD code in a page's <head> declaring 'this page has Q&A pairs'. One of the strongest single GEO levers — our current observations suggest pages that add it are cited noticeably more once the page also answers the question directly.
- HowTo schema
- JSON-LD declaring 'this is step-by-step instructions'. AI engines extract clean ordered steps from this.
- Article schema
- JSON-LD with author + publisher + date. Strengthens E-E-A-T signals so engines trust your content more.
- Entity JSON-LD
- Organization/brand schema that tells AI engines who you are, with sameAs links to your verified profiles (LinkedIn / Crunchbase / Wikipedia).
- E-E-A-T
- Experience, Expertise, Authoritativeness, Trust — the trust signals AI engines use to decide who to cite.
- Readiness score
- Your 0-100 grade on the 7 AI-readiness checks.
Questions, answered
Site Audit FAQs
Which AI engines are tracked?
Why does the banner say DEMO MODE?
How do I flip the module to real-data mode?
What's the difference between 'cited' and 'mentioned'?
What is the Brand Authority Graph?
What is AI Snippet Preview?
How does competitor benchmarking work?
What are the 5 moats no competitor ships?
Can I use my own customer questions instead of generic ones?
What's the AEO rewriter for?
Can I deploy llms.txt + entity JSON-LD straight to my site?
Will I get alerted if my citation rate drops?
How often should I run citation checks?
Is my citation history accurate after I flip from demo to real mode?
What's the difference between the Citations tab and the Insights tab?
How does the cross-engine confidence badge work?
How is URL Discovery different from prompt discovery?
How we compare
Honest comparison — where we win, where the big tools win.
vs. Profound ($499/mo)
We ship every column they ship (citation tracking, sentiment, quality score) PLUS the columns they don't (topic clusters, AI traffic attribution, Brand Authority Graph, schemas + AEO rewriter) — at Pro+ pricing (~$29 included, not a $499 add-on).
vs. Athena HQ ($249/mo)
Their pitch is 'AI Search Console' (AI traffic attribution). We ship that as Moat 4 + the other 4 moats they don't (sentiment, clusters, quality score, authority graph) + 6 engines tracked vs their 3.
vs. Otterly ($99/mo) / Peec ($89/mo)
They do citation tracking on 3-4 engines and stop there. We track 6 engines (added the AI, Google AI Overviews, Grok) + 5 deeper moats + schemas + AEO rewriter — all included.
vs. Ahrefs / Semrush
They sell GEO add-ons starting at $39/mo extra, and they only do citation tracking. We built GEO as a first-class module — citation tracker + 5 moats + FAQ/HowTo/Article schema generators + AEO content rewriter in one place.
vs. DIY (ChatGPT + Schema Markup Generator + manual checks)
You'd spend hours per page. Outerank does it in seconds, with real citation data + Brand Authority Graph + sentiment + competitor benchmarks + a content rewriter trained for AI-quotability.
Ready to try AI Search Optimizer?
Open it on one of your projects.
Come back here whenever you need a refresher.