GEO starts with good SEO
GEO (generative engine optimization) means making a page easy for an answer engine to use: Google’s AI Overviews and AI Mode, ChatGPT, Perplexity, Claude. These engines rely on web pages. They pick the ones that are accessible, answer the question clearly and can be trusted.
So LLM SEO doesn’t replace SEO, it builds on it. A page blocked by noindex or nosnippet won’t be cited anywhere. An answer buried under three paragraphs of introduction is unlikely to be picked up. What changes is the scale: an AI quotes a passage, not a whole page.
Nobody can guarantee a citation in ChatGPT or an AI Overview, and Searkin doesn’t promise one. What you can control is the quality and completeness of each page. That’s what the diagnosis measures.
The eight-criteria diagnosis
Calculated at every audit, with no AI and no credits. Each criterion gets a status, each issue an explanation and an action.
- A
Main answer
The question the page answers, and whether the answer comes quickly or gets lost after a long introduction.
- B
Heading structure
Missing or multiple H1, skipped levels, duplicate headings, and topics real searches expect but the outline doesn’t cover.
- C
Originality
Generic passages, and what only this site can say: examples, figures, client cases.
- D
Expertise and sources
Identifiable author, publication and update dates, links to sources, claims that need one.
- E
Internal linking
Orphan page, article attached to no cluster, missing link to its pillar page.
- F
Schema.org
Unreadable JSON-LD, missing properties (headline, author, date), a type that doesn’t match the page.
- G
Google access
Noindex, robots.txt, canonical to another page,
nosnippet, content that only exists after JavaScript. The only criterion that can be “blocked”. - H
Images and video
Images without alt text, long content with no visuals, a diagram or table that would help.
Where to start: priority pages
On a 300-page site, almost every page has something to improve. The AI answers tab ranks them for you by combining four signals:
- a blocked page comes before everything else;
- potential: the clicks it would gain at the click-through rate expected for its position;
- the share of question-type searches (“how”, “why”, “what”…), the ones most likely to trigger an AI answer;
- the drop in its clicks over the last 28 days.
Worth knowing: Search Console counts AI Overview impressions together with classic results, without separating them. Searkin says so on screen instead of making up a number.
AI review, on demand
For what rules can’t measure, the AI reads the page alongside its real Search Console queries.
You pick the pages
1 to 20 pages at a time, from the AI answers tab. The credit cost shows first, lower if the AI runs on your Claude subscription.
The AI reads the page and its searches
Main content, heading outline, the page’s top 30 Search Console queries and the site’s editorial guidelines.
It makes the call
Main question, clear, partial or missing answer, generic passages, claims to source, expected topics that are missing.
Nothing made up
Every quote must appear word for word on the page, every query in Search Console. Anything else is removed before display.
Recommendations followed through
- Sorted by priority: high (blocked page, missing answer, unsourced claim), medium, low.
- For each: the issue, why it matters, the suggested action and the heading concerned.
- A status for each recommendation: to do, done or dismissed.
- A fix marked as done is checked again at the next audit, and reopened if the issue is still there.
- No FAQ added by reflex, no minimum word count: that kind of hollow advice is filtered out.
- Everything is also available from Claude, through the MCP connector.
AI Overview tracker for your keywords
The diagnosis measures page quality. To know whether a page is actually cited, you have to look at the results page. Tick “Also record AI Overviews” in rank tracking: for each tracked keyword, Searkin records every day whether an AI Overview appears, whether the site is cited, which page is cited, and which sources are cited instead.
It’s included in real rank tracking (top 100) at no extra cost. For now, tracking covers Google France desktop results only.
GEO: frequently asked questions
Does Searkin guarantee a place in ChatGPT or AI Overviews?
No, and nobody can. Searkin measures what’s in your hands: an accessible page, a clear answer, sources, a readable structure. Then it checks, for your tracked keywords, whether an AI Overview actually cites the site.
What’s the difference between GEO and SEO?
SEO aims to rank a page in the results. GEO aims to get it used in an AI-generated answer. The foundations are the same; GEO focuses on the snippet: a clean answer, in the right place, that an AI can quote as is.
Should I add an FAQ or special markup?
Not by default. No markup guarantees a place in AI Overviews, and an FAQ added for no reason weighs the page down. Searkin deliberately leaves that advice out and prefers to put the answer in the right place.
What about ChatGPT, Perplexity or Claude?
Searkin currently tracks Google’s AI Overviews, the only AI answer that can be checked daily at scale. The diagnosis criteria apply to every answer engine: they also rely on pages that are accessible, clear and sourced.
Does Searkin change my pages?
The GEO diagnosis recommends; it writes nothing to the site. Titles, descriptions and redirects, on the other hand, can be written into WordPress or Payload once you approve them.
How much does the GEO diagnosis cost?
The eight criteria are included in the site price. AI review is paid in credits, at a lower cost with your Claude subscription. AI Overview tracking is included in rank tracking. Details are on the pricing page.