SEO / AEO / GEO Agent
Continuous content optimization for search engines, AI answer engines, and generative search, in any industry
This agent is live, try it right now on your own page.
Paste any public page URL, any industry, and get back a structured SEO/AEO/GEO audit in under a minute.
Audit any page on your site
Paste a single URL: a product page, a pricing page, a clinic or store location, a service page, a job posting, a news article, anything public in any industry. The agent reads the sector off the page first, then scores SEO/AEO/GEO findings against that sector's conventions, with verbatim evidence and specific fixes. The full report is downloadable as Word, Markdown, or JSON.
For the best experience, launch the demo in its own tab.
The inline preview below can show up blank if your browser blocks embedded apps (common in Safari, Brave, Firefox, and Chrome with tracking protection). Launching opens the full agent in a new tab where it always runs, if it's been idle, it wakes in about 10 seconds.
Preview blank or stuck loading? Your browser is blocking the embedded session: launch it in a new tab instead, or .
How you get your audit
- 1. Paste any public URL on your domain, in any industry.
- 2. Audit renders in 20–60 seconds: detected sector, scored findings, fixes, AI visibility.
- 3. Download as Word, Markdown, or JSON, saved directly to your device.
No account, no email needed for the demo. Production deployments run continuously across your full site and email weekly briefs.
Deploy this agent
Full-domain crawl, SERP + AI-engine probes, content briefs to your CMS. Zero upfront cost, pay only for measurable, pre-agreed results.
Book a 20-minute callOverview
The SEO / AEO / GEO Agent monitors how your organization surfaces across traditional search (Google, Bing), AI answer engines (Google AI Overviews, Bing Copilot), and generative platforms (ChatGPT, Perplexity, Gemini, Claude). It works in any industry: it identifies the sector a page belongs to and scores it against that sector's conventions, so a product page is measured on product markup, a clinic or trades page on location, hours and service area, a SaaS page on software and pricing markup, a careers page on job markup, a school page on course and program markup. It finds which pages are losing visibility, which queries you are being misrepresented on, and where competitors are winning citations you should own. Every week it generates prioritized content updates, new FAQ blocks, refreshed page copy, schema additions, internal linking changes, that your web team can publish in minutes.
Integrations
Connects to Google Search Console, Bing Webmaster, GA4, Screaming Frog crawls, your CMS or storefront (WordPress, Webflow, Shopify, HubSpot, Contentful, Drupal, Sitecore, Cascade), Google Business Profile for multi-location businesses, and the major AI search APIs for citation tracking.
Key Metrics Tracked
- AI engine citation rate for priority queries
- Organic visibility on high-intent commercial terms
- Content updates shipped per week
- Click-through and enquiry or purchase conversion lift
The nine checks every engagement starts with
Before a word of content is written, the agent reads your site the way an AI assistant reads it and scores nine things that decide whether ChatGPT, Gemini or Perplexity can use you in an answer at all. These nine are sector-independent: a manufacturer, a hospital, a law firm and a university fail and pass them the same way, and only the markup that satisfies check two changes with the industry. Every verdict comes from a live fetch of your pages. Nothing on the scorecard is typed by hand.
1. A description of your organization that machines can read
A plain statement in the page's code of what the organization is, where it is and what it offers. AI assistants use it in preference to guessing from page text.
2. What you sell, marked up as what it is
Markup that names the thing: a product, a service, a treatment, a job opening, a degree program. The right type differs by industry, and the wrong type is worth about as much as none. Without it, a product name is just words in a sentence to a machine reading the page.
3. Common questions in a form that can be quoted
AI answers are built from short, direct answers to real questions. Price, availability, coverage area, lead time and eligibility are the ones buyers ask, whatever the industry.
4. A guide file for AI assistants
A short file at the root of the site telling AI tools which pages matter. New, cheap, and most sites do not have one yet.
5. AI assistants allowed to read the site
The retrieval crawlers that fetch a page to ground a live answer. Plenty of sites have switched these off and are now missing from those answers entirely.
6. A map of the site AI crawlers can find
The file that lists every page you want found. If it is not declared where crawlers look, they work from whatever links they stumble across.
7. That map working at the standard address
The one address every crawler tries first.
8. A homepage summary
The one or two sentences that describe the organization. It is what a search engine and an AI assistant read first.
9. A main heading on the homepage
The single line that says what this page is. It anchors everything a machine understands about the page.
Training crawlers and retrieval crawlers are scored separately
Most audits flatten every AI user-agent into one list, which produces a finding that is simply untrue. The two classes do different jobs and blocking them costs different things.
- Retrieval (OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, Claude-SearchBot, Claude-User): fetches a page to ground a live answer. Blocking one removes you from that engine's answers. We score it.
- Training (GPTBot, ClaudeBot, CCBot, Bytespider, Applebot-Extended, Meta-ExternalAgent): corpus collection for model training. Blocking is a defensible publisher policy and costs nothing in AI answer visibility. We report it and leave it unscored.
- Both (Google-Extended): gates Gemini training and Gemini's grounded answers. Blocking it costs live Gemini visibility, and it does not affect Google Search or AI Overviews, which use Googlebot.
So “you block GPTBot, therefore ChatGPT cannot see you” is false, and you will never see it in an Ibex report. Your web team knows the difference, and a finding that fails their check fails the whole document.
What to do about it, in the order it should be done
The fix list is derived from which checks failed, so a site that already has a homepage heading is never told to add one. Sequence matters more than volume here. Crawler access first, because nothing downstream is read until it is fixed. The AI guide file last, because it does nothing until the pages it points at are in order.
1. Reopen the site to the assistants that answer questions
One line in robots.txt. Google-Extended gates Gemini's answers as well as its training data, so that block costs visibility today. The crawlers that serve ChatGPT search are a separate list and can be decided separately.
2. Give the homepage one main heading and a real summary
An hour of work. It is the first thing every engine reads, and on most sites we screen, in every industry, it is empty, nearly empty, or spread across so many headings that none of them leads.
3. Add machine-readable detail for the organization and every page that sells something
A week or two, using the schema types your sector actually uses: Product and Offer for retail, a LocalBusiness subtype with hours and service area for a clinic or a trade, SoftwareApplication for software, JobPosting for careers, Course and program types for education. Nothing visible changes for a visitor. The page starts stating what it is instead of implying it.
4. Publish the site map where crawlers look, and add the AI guide file
An afternoon, once the pages themselves are in order.
5. Answer the questions buyers actually ask, on the pages that sell
Ongoing. Price, availability, coverage, lead time, eligibility, warranty, in short direct answers. This is the material AI answers are assembled from, and the questions change with the industry even though the format does not.
What the scorecard will not claim
It reports what your site hands a reader and what that costs. It does not claim you are missing from a specific AI answer unless that citation test has actually been run against your query set, which is the tracked prompt set in a deployment rather than an assumption on a first-pass screen.
What deployment actually looks like
What you see on this page is a demo. Sample data. The shape of the work.
Agents earn their keep when they orchestrate actions on your data. That is what we build.
We take your proprietary data and wire up the automations and AI agents that run alongside your team.
End-to-end automated, or collaborative. You stay in the loop on the actions that matter. The work happens privately, inside your systems.
Frequently Asked Questions
How is this different from a normal SEO agency?
How do you measure AI engine visibility?
Will it write content automatically?
Does blocking GPTBot hurt our AI visibility?
What is llms.txt, and do we need one?
Does this only work for higher education?
What do you check before writing any content?
Disclosures, data handling & legal notices
Read this before using the live demo above. By running an audit you accept the terms summarized here. Last updated 2026-09-09.
- The audit is generated by a third-party AI model (US-based provider). Output may contain errors; verify findings before acting on them.
- Only audit pages that you own or have permission to audit. The demo respects
robots.txt; you should too. - The page you submit is fetched once and sent to the AI provider for analysis. Ibex does not store the URL, the page contents, or the audit report. The AI provider may retain inputs per its API terms.
- Do not submit pages containing personal data (customer or student PII, patient records, internal-only data). The demo is for public marketing pages, in any industry.
- This is informational, not professional SEO, legal, marketing, or compliance advice.
- Production deployments run inside your own security boundary, accommodate DPAs / BAAs / security questionnaires, and are built to support the regime your sector answers to (FERPA for education, HIPAA for healthcare, and so on). The public demo does none of that.
1. AI-generated content & accuracy
The audit report is generated end-to-end by a third-party large-language model running on a US-based inference platform. Page fetch, signal extraction, and report rendering are deterministic; the analytical findings, scores, recommended fixes, and AI-visibility paragraph are model-generated.
Like all current LLMs, the model may produce: factually inaccurate findings; recommendations that are technically misguided; severity ratings that do not match real-world impact; misattributed evidence (the model may quote text from elsewhere on the page rather than the section it intends to cite); confident-sounding statements that are wrong (“hallucinations”); biased output reflecting biases in the model's training data.
One failure mode is specific to this agent: it infers your industry from the page before scoring, so a thin or ambiguous page can be read as the wrong sector and then measured against the wrong conventions. The report prints the sector it inferred and its confidence for exactly this reason, and the sidebar industry field overrides it.
Verify every finding against the actual page before acting. Do not present demo output as a definitive technical assessment, an accreditation-ready audit, or a substitute for a qualified SEO/marketing engagement. The demo is best used as a fast first-pass screen for obvious issues and as a thinking partner, not as a final answer.
2. Data handling & sub-processors
What the demo collects: the URL you submit, the audit-context fields you optionally fill in (organization name, industry, page goal, target audience, target queries), the rendered HTML and extracted text of the page at that URL, and standard server-side request metadata (timestamp, IP address, logged only as part of normal Streamlit Cloud platform logging).
What Ibex stores: nothing. The URL, the fetched HTML, the extracted signals, and the generated audit report are processed in memory and discarded when your browser session ends or the demo container restarts. Ibex does not persist demo inputs or outputs to a database.
Sub-processors involved in serving the demo:
- Streamlit Cloud (Snowflake Inc.) – hosts the demo container at
ibex-seo.streamlit.app. Sees the URL you submit, receives the audit request, and returns the rendered report. Streamlit's privacy terms apply to platform telemetry and access logs. - Third-party AI inference provider (US-based) processes the inference call. Receives the system prompt, the audit-context fields, the extracted page signals, and up to 8,000 characters of the page's body text. Per the provider's API terms, inputs may be retained for service-quality and abuse-prevention purposes. The provider's name and current data-retention policy are available on request to team@ibexinsights.co.
- GitHub, Inc. – hosts the source code (private repository) but does not see demo traffic.
- Netlify, Inc. – hosts the marketing site at
ibexinsights.co; receives standard request logs for the marketing pages, including the page that embeds this demo.
What is NOT collected: no accounts, no email addresses, no payment information, no third-party tracking pixels (no Google Analytics, no Meta Pixel, no LinkedIn Insight Tag, no session-replay tools).
If you submit personal data accidentally: contact team@ibexinsights.co. Note that for the demo, Ibex itself does not retain your data; you may also need to contact the sub-processors above.
3. Acceptable use
By using the demo you agree that you will:
- Only audit pages you own, administer, or have explicit permission to audit. Auditing a competitor's page or a third-party site you do not control may violate that site's terms of service or applicable computer-misuse / anti-scraping laws (e.g., the U.S. Computer Fraud and Abuse Act, the U.K. Computer Misuse Act, similar statutes elsewhere). You are solely responsible for the legality of any URL you submit.
- Respect
robots.txt, paywalls, and authentication. The demo fetches with a transparentUser-Agentidentifying it asIbexInsightsAuditBot/1.0; the demo does not attempt to bypass authentication, CAPTCHAs, JavaScript-only content, or paywalls. Do not use the tool against pages that require login or are explicitly disallowed by theirrobots.txt. - Not use the demo for bulk or automated workloads. The demo is rate-limited at the inference provider (free-tier capacity shared across all visitors). Programmatic / scripted use is prohibited.
- Not submit pages containing personal data – student personally-identifiable information protected by FERPA, protected health information under HIPAA, EU/UK personal data under GDPR, California consumer data under CCPA, child data under COPPA, or any other regulated personal data. The demo is intended for public marketing pages.
- Not attempt to reverse-engineer, evade limits, prompt-inject, or extract the system prompt.
- Not use audit output to harass, defame, or misrepresent third parties. “Site X has terrible SEO” based on a demo run is not a defensible public claim.
Ibex reserves the right to rate-limit, block, or restrict access from IP addresses or networks that violate these terms.
4. Regulatory posture, demo vs. production
The public demo and production deployments differ materially in their regulatory posture. Do not treat the demo's data path as representative of a contracted engagement. The frameworks below are the ones that come up most often; which of them binds a given deployment depends on your sector and is settled in the contract, not here.
| Framework | Public demo | Production deployment |
|---|---|---|
| FERPA (20 U.S.C. § 1232g) | Not for student records. No FERPA representation. | For education customers: built to support FERPA compliance; runs inside the customer's perimeter; Ibex executes a Data Processing Addendum where required and never uses customer data for model training. Customers outside education are unaffected by FERPA. |
| GDPR / UK GDPR | No personal data should be submitted. EEA/UK visitors should treat the demo as a public, opt-in tool with US-based sub-processors. | DPA executed; standard contractual clauses; data residency options available; GDPR Article 28 sub-processor obligations honored. |
| HIPAA | Not a HIPAA-eligible service. Do not submit PHI. | BAA available for customers handling protected health information: healthcare providers, payers, employer health plans, and campus health services. |
| Vendor security review (HECVAT, SIG, CAIQ, VSA) | Demo is not a scoped service under any of these questionnaires. | HECVAT Lite for education procurement; SIG Lite, CAIQ, or a customer questionnaire for every other sector, completed on request. |
| SOC 2 / ISO 27001 | Demo runs on Streamlit Cloud (SOC 2 Type 2 attested) and a third-party AI inference provider (attestation status per provider). | Production roadmap: SOC 2 Type 2 against the Ibex platform itself. |
| CCPA / CPRA (California) | No sale of personal information. The demo does not set tracking cookies. California residents may request information at team@ibexinsights.co. | Service Provider terms in MSA; CCPA-compliant data subject request handling. |
| TCPA / CAN-SPAM | Demo sends no marketing communications, no email/SMS collection. | Outreach-related agents include opt-in capture, opt-out management, and per-jurisdiction consent rules. |
| ADA / WCAG 2.1 AA | This page targets WCAG 2.1 AA. Report accessibility issues to team@ibexinsights.co. | Production interfaces target WCAG 2.1 AA; VPATs available on request. |
| EU AI Act / state AI laws | Demo is a general-purpose informational tool; no high-risk decisions are automated. Output is advisory only. | Risk classification per use case; agents touching hiring, lending, insurance, enrollment, or other consequential decisions are scoped, logged, and human-reviewed in line with applicable AI-governance frameworks (NIST AI RMF, EU AI Act risk tiering, Colorado AI Act, NYC Local Law 144 where relevant). |
| Section 508 (federal) | Best-effort accessibility on the demo. | Section 508 conformance available for federal, public-sector, and federally-funded customers. |
| Sector-specific regimes (GLBA, PCI DSS, FINRA, state bar advertising rules, FTC endorsement guides) | Out of scope. The demo processes a public marketing page and nothing else. Do not submit regulated records. | Scoped per engagement in the MSA and DPA. The agent's function is page content and markup, so it is not a system of record for regulated data; where published claims are regulated (financial promotions, legal advertising, health claims), drafted copy goes to your compliance reviewer before publication like any other content. |
| U.S. export controls (EAR) | The demo is provided from the United States. Visitors in U.S.-embargoed countries (per OFAC) should not access the service. | Customer-side compliance is the customer's responsibility; Ibex does not transfer controlled technology. |
Compliance claims about production deployments above are forward-looking targets for contracted engagements; the controls applicable to a specific contract are governed by the executed MSA / DPA / BAA, not by this page.
5. Intellectual property
The page you submit: the content of the URL you audit belongs to its owner. The demo fetches a single copy for analysis and does not redistribute, cache publicly, or republish that content. Quotes that appear in audit findings as “evidence” are short verbatim excerpts used for the limited purpose of supporting an analytical claim, the kind of fair-use / fair-dealing excerpting an SEO consultant or a journalist would do.
The audit you receive: you may use, share, and republish the audit output your demo run produced, including in internal documents, vendor evaluations, board materials, or public reporting about your own pages. Ibex does not assert copyright over your audit output.
Ibex marks: “Ibex Insights”, the Ibex word and logo marks, and the Ibex agent product names are trademarks of Ibex Insights. The marks may be used in factual references (“we used the Ibex SEO/AEO/GEO Agent demo to…”) without permission. Use of the marks in a way that suggests endorsement or partnership requires written consent.
Third-party model output: the underlying language-model output is generated by a third-party AI provider whose terms govern allowed uses of model output. For the audit's informational purpose this is a non-issue, but commercial redistribution of bulk model output may be subject to the provider's license terms.
6. No warranty · limitation of liability
The demo is provided “as is” and “as available”.
To the maximum extent permitted by applicable law, Ibex Insights makes no representation or warranty of any kind, express or implied, including without limitation: warranties of merchantability, fitness for a particular purpose, non-infringement, accuracy, completeness, timeliness, uninterrupted availability, or that the audit will achieve any particular ranking, citation, or conversion outcome.
The demo is offered free of charge as a marketing tool. Ibex Insights, its officers, employees, and contractors will not be liable for any indirect, incidental, special, consequential, exemplary, or punitive damages, or any loss of profits, revenue, data, goodwill, or business opportunity, arising out of or in connection with use of the demo, even if advised of the possibility of such damages. To the extent any liability is found notwithstanding this disclaimer, total aggregate liability shall not exceed one hundred U.S. dollars (US$100) per claim.
Some jurisdictions do not allow the exclusion or limitation of certain warranties or liabilities; in those jurisdictions, the above limitations apply to the maximum extent permitted.
7. Demo availability · rate limits · changes
The demo runs on a free hosting tier and may be unavailable, slow, or rate-limited from time to time. There is no service-level agreement (SLA) for the public demo. The demo container may sleep after periods of inactivity and take a few seconds to wake.
Ibex may modify, suspend, or discontinue the demo at any time without notice. Material changes to these disclosures will be reflected in the “Last updated” date at the top of this section.
The shared inference quota at the AI provider is finite. Heavy concurrent usage may temporarily prevent new audits; please retry after a minute.
8. Children · COPPA
The demo is intended for marketing, communications, web, and IT professionals in any industry. It is not directed at and not intended to be used by children under the age of 13. Ibex does not knowingly collect data from children under 13. If you believe a child under 13 has interacted with the demo, contact team@ibexinsights.co.
9. Governing law · disputes
These disclosures and any dispute arising from use of the demo are governed by the laws of the jurisdiction in which Ibex Insights is established, without regard to conflict-of-laws principles. The parties agree to attempt resolution in good faith via direct correspondence before initiating any formal proceeding.
Contracted engagements are governed by the executed Master Services Agreement, Data Processing Addendum, Business Associate Agreement (where applicable), and statement of work, which supersede this section for the engagement's scope.
10. Contact · legal · security disclosure
General & legal inquiries: team@ibexinsights.co
Data subject requests (GDPR / UK GDPR / CCPA / CPRA): team@ibexinsights.co with subject line “Data subject request”.
Accessibility issues: team@ibexinsights.co
Responsible security disclosure: team@ibexinsights.co. Please include reproduction steps; we will acknowledge within 5 business days. Do not exploit any finding beyond what is necessary to demonstrate the issue, and do not access data that is not your own.
DMCA / IP takedown: team@ibexinsights.co
These disclosures cover only the public demo embedded on this page. They do not constitute a Master Services Agreement, Data Processing Addendum, Business Associate Agreement, or other contract; engagements with companies, universities, and other organizations are governed by separately executed agreements. Last updated 2026-09-09.