Why GEO
People increasingly ask AI engines directly instead of running a search. Showing up — and showing up well — inside those AI answers is its own discipline. GEO measures how often AI engines mention you, how prominently, whether they cite you, and whether they recommend you over competitors. It then helps you fix the gaps.Where to find it
- Open the client and select the active site.
- Expand Channels and select GEO.
- Score — your visibility metrics across AI engines (the default view).
- Prompts — tracked questions and prompts being evaluated for future tracking.
- Fanout — the searches and sub-questions engines used to build answers.
- Citations — where you’re cited across engines, and the traffic that follows.
- AI Sentiment — the specific claims and language AI engines use about your brand.
- Competitors — how you compare against configured competitors.
- Analytics — referral, search, and performance evidence from connected sources.
- Improvements — prioritized recommendations for raising AI visibility.
- Foundational — public facts, surfaces, and training-data evidence for longer-term model visibility.
What you see on each sub-tab
Score (the headline view)
The default view. Cards here include:- AI Visibility Score Trend — a chart of your visibility over time, with a time-range selector.
- GEO Score — your composite score, with per-engine breakdowns.
- How Visible Are You To AI? — per-engine visibility tiles (ChatGPT, Perplexity, Google AI Overview, Claude, etc.) showing the visibility score and the run-over-run change.
- Sentiment Score — overall tone of how you’re described, with a quality / service / price breakdown.
- Citation Drift — whether your citations are stable or changing run to run, with domain change and trend.
- Measurement Precision — sample size, the minimum needed, and a reliability label.
Cleo Audit
Cleo Audit is Cleo’s all-in-one audit of how ready your site is for AI engines and search. It now lives under SEO → Audit, while its findings remain directly relevant to GEO. It runs a crawl and reports:- A Cleo Audit Score (out of 100) with a Good / Fair / Poor label, plus headline stats: Issues Detected, Pages Crawled, and Errors.
- Category scores across the things that decide whether AI engines can read and trust your site — for example Crawler Access, JS Rendering, Structured Data, Content Quality, Content Size, and
llms.txt& Meta, plus deeper categories like AI Search Presence, Links & Navigation, and Performance (Core Web Vitals). - A Deep Insights tab — a single issues table with every finding, grouped by category. Rows expand to show the affected pages, the impact, and a step-by-step fix.
- A Pages tab — the pages that were crawled.
- A Robots & Sitemap tab —
robots.txtand sitemap validation.
Citations
The Citations sub-tab shows where AI engines are pulling their answers from:- Citation Share — your slice of the sources engines cite for your tracked queries.
- The Visibility Funnel — how often you’re Mentioned, Cited, and Recommended.
- Citation Trajectory — the top sources cited across engines over time, with a date-range selector.
- Presence by Engine — which engines surface you and how often.
- Query prompts — the high-intent prompts where citations matter most.
Competitors
Side-by-side against the competitors you’ve configured:- Who is mentioned more often across your tracked queries.
- Who appears earlier in answers.
- Who is framed as a recommendation more often.
- Where each competitor wins and where you do, with a trend view over recent weeks.
Analytics
The Analytics sub-tab (previously “Results”) measures real-world impact: traffic and branded search. It connects to:- Google Analytics 4 (GA4) — to track AI referral traffic (visits coming from
chatgpt.com,perplexity.ai, etc.). - Google Search Console (GSC) — to track branded-search trends.
Search Console visibility
- Change against the previous period. The branded queries and AI-cited pages tables compare your selected range with the same number of days just before it. A sync fetches both periods, so the comparison appears after the first sync.
- Zero-click pressure. A row is flagged when impressions held steady or rose, clicks fell, and average position moved less than one place. An AI Overview answering the search on the results page often looks like this, but so do seasonal swings, a change to Google’s results layout, or a new competitor. Treat the flag as a prompt to look closer. It is not a measurement of traffic lost to AI Overviews.
- Ranked by Google, not cited by AI. Pages already earning Google impressions that none of your tracked AI engines cite. These are the clearest content gaps to work on. Cleo keeps the busiest 250 such pages with at least 10 impressions over the synced range.
- Generative AI report. The link in the Search Console header opens Google’s own Generative AI report (beta) for your property, which counts impressions inside AI Overviews and AI Mode. Google has no API for this report yet, so Cleo cannot show the numbers itself. If Google hasn’t enabled the report for your property, or it doesn’t have enough AI feature impressions, the report won’t appear. That means Google has no data for your property yet, not that you have zero impressions.
Connect GA4 and GSC
- Open the Analytics sub-tab.
- Authorise Google when prompted.
- Select the GA4 property and GSC site for this project.
- Save.
Improvements
This is where you go to raise your GEO score. The Improvements sub-tab analyses your tracked queries and surfaces concrete, prioritised recommendations — Cleo tells you what to fix and which tool to use. Recommendations typically cover:- Content to create — topics and questions where AI engines cite competitors but not you.
- Pages to improve — existing pages that are close to being cited but need strengthening.
- On-page fixes — structural changes that make content easier for AI engines to read and cite: FAQ sections, schema, comparison sections, citation-friendly blocks.
- Tool suggestions — for each recommendation, Cleo points to the specific SEO tool to use.
- Open the site workspace and choose GEO → Improvements.
- Review the prioritised list.
- Click a suggestion to see the detail and the recommended tool or action.
- Use the suggested tool directly, send it to Tasks for a teammate, or send a content suggestion to Quill to draft the article.
Foundational
Everything else on the GEO tab measures retrieval: an AI engine searching the web at answer time and finding you. Foundational is about the slower, deeper layer: whether the next generation of models learns your brand during training, so it can describe and recommend you without searching. Be clear about what this can and cannot do:- Training data is collected months before a model ships, so nothing you do here shows up in a model for 6 to 18 months, and only when a vendor releases a new model.
- A fact needs to appear many times, in different wording, across sites the model builders trust, before it is represented reliably in training data. One page is not enough.
- Nobody can prove a closed model trained on a given page. Cleo shows the strongest honest proxies instead.
- Fact sheet. The canonical facts about your company. Your name, aliases, description, category, country and products come straight from your GEO settings, so you never type them twice and the two can never disagree. Add the few things GEO settings has nowhere to put: legal name, headquarters city, founding year and founders. Your official profile URLs (YouTube, LinkedIn, X, Wikidata, company registry) are found automatically from your own site where you link them, so you usually need to add nothing. Every check below derives from these, so keep them accurate.
- Training crawlers. Verified visits from AI training crawlers (GPTBot, ClaudeBot, CCBot and others) to your site, fed by the Cleo WordPress plugin or CloudFront. This moved here from Analytics; AI-answer and indexing crawlers stay on Analytics.
- Foundation readiness. Run an audit to check four things: whether training crawlers are allowed (robots.txt and edge blocks such as Cloudflare), whether your key pages are discoverable and server-rendered, whether your About and product pages would survive the public quality filters used to build AI training sets, and whether your facts are consistent across your site and Wikidata.
- Surfaces. Where your brand is present on sources that demonstrably reach AI training data (Wikidata, Wikipedia, the Google Knowledge Graph, YouTube and your company registry entry), and a separate list of surfaces that help AI search but not training (Crunchbase, G2, Trustpilot, LinkedIn, llms.txt, schema markup) so you are not sold the wrong lever.
- Training data evidence. How many of your pages the public web archive captured on its latest crawl, plus how often your brand name appears in the public datasets open AI models are trained on. Refreshed monthly.
- Actions. A ranked list of what to do, each labelled with how strong the evidence is, how long it takes to have an effect, and how Cleo will verify it. Some open the relevant fix tool, some draft text for you (About-page paragraphs in several phrasings, a robots.txt block, Wikidata statements with references, a Wikipedia edit request that follows the conflict-of-interest rules).
Set up GEO for a site (onboarding)
The first time you open the GEO tab for a site that has no queries yet, Cleo runs a short onboarding flow:- Industry — confirm your industry and business description so Cleo understands your context.
- Queries — Cleo analyses your site and proposes tracked queries; you review, edit, add, or remove them.
Add, edit, or remove queries
Unlike older versions of Cleo, queries are managed inside the app at any time — you don’t need to contact support to change them:- Click Add Query on the GEO tab to add new queries (Cleo can suggest some for you). You can add up to your plan’s query limit.
- Use the edit icon on a query row to change its wording, or remove a query you no longer want to track.
Add a competitor
Add, edit, or remove competitors on the Competitors sub-tab, up to your plan’s competitor limit.GEO settings
The gear icon opens GEO settings, where you can update your Brand Name, Industry, and Business Description, and see which AI engines are enabled for your plan.Run cadence
How often a query runs depends on where it sits in its lifecycle, not on your plan. Runs happen Monday–Saturday:
The queries you are still contesting are checked most often. A query you have won consistently drops to a weekly check, because a question you already own does not need watching daily — and if it starts slipping, four losing weeks move it back to In Play and daily checks resume automatically.
Your site’s weekly and twice-weekly checks all land on the same day each week, so you get one complete refresh rather than a trickle. Cleo aggregates these runs over time to produce stable trends. What differs by plan is how many queries, engines, and competitors you can track.
The GEO metrics
Cleo’s full eight-metric GEO suite:- AI Presence Rate — how often any AI engine mentions you in answers to your tracked queries.
- Answer Prominence — how high in the answer your mention appears.
- Citation Rank Index — your ranking in the citations the AI provides.
- Share of Voice — your share of brand mentions vs competitors across your queries.
- Recommendation Rate — how often you’re framed as a recommendation rather than just a mention.
- Sentiment Score — the tone of how you’re described (with quality / service / price breakdown where available).
- Citation Drift — whether your citations are stable or changing run to run.
- Confidence Interval / Measurement Precision — the statistical confidence around the numbers above (chiefly a function of how many queries you track and how much daily data has accumulated).
Reading the numbers honestly
AI engines are non-deterministic. The same query can return different answers run to run. That’s why Cleo:- Runs prompts on their configured lifecycle cadence and aggregates the results over time.
- Reports a measurement precision / confidence interval alongside the headline numbers.
- Tightens that confidence as more data accumulates and as you track a broader useful prompt set.
Plan limits
The number of prompts, engines, and competitors depends on your plan. Run cadence follows each prompt’s lifecycle stage (see Run cadence above). If you hit a limit, the dashboard explains the available next step. See Plans and billing.Troubleshooting
- No data on Score: finish the GEO onboarding flow, add at least one prompt, and wait for the first run.
- I want to add or change a query: do it in-app — click Add Query, or use the edit icon on a query row. You can track up to your plan’s query limit.
- No data on Analytics: finish the GA4 and GSC connection on the Analytics sub-tab, then wait for the first sync.
- A query keeps failing: the engine may be slow or temporarily unavailable; re-run later.
- Mention count looks wrong: the AI may be referring to you indirectly. If it’s a genuine parsing issue, write to contact@regencleo.ai.