Straight answers to the questions people ask about being found on Google and cited by AI assistants, with examples you can check yourself and the limits stated plainly.
How assistants pick their sources, what blocks a site, and what you can reasonably expect.
GEO (Generative Engine Optimization) means making your site easy to read, understand and cite for AI assistants such as ChatGPT, Perplexity, Gemini, Claude or Copilot.
SEO gets you found on Google. GEO gets you cited in an assistant’s answer. Both are worked on the same pages: a site robots can read, clear answers from the first lines, precise and sourced facts.
The GEO guide explains the method in 7 steps, with the sources of the studies it quotes.
For a recent or specific question, ChatGPT searches the web, reads a few pages, then cites the ones that answer best. Without a search, it answers from what it learned, and your brand’s reputation is what counts.
ChatGPT’s search relies heavily on Bing’s index, completed by its own robot, OAI-SearchBot. A page missing from Bing has little chance of being cited.
ChatGPT quotes passages, not whole pages: a paragraph that clearly answers a question is more likely to be picked up than a long text that circles around it.
Limit : Studies on the topic are correlational. They show what cited sites have in common, without guaranteeing that a given action will lead to a citation.
Check that your robots.txt and your server let through OAI-SearchBot (ChatGPT’s search) and ChatGPT-User (ChatGPT opening a page when a user asks).
Many sites block them without knowing: a rule that’s too broad in robots.txt, a security plugin, or the host’s anti-bot protection.
Our AI robots checker reads your robots.txt and tests each robot’s real access, free and without an account.
Check it yourself : a robots.txt open to ChatGPT’s search
Open mysite.com/robots.txt in your browser. No Disallow: / line should target these robots, nor the User-agent: * group they follow by default.
User-agent: OAI-SearchBot
Allow: /
User-agent: ChatGPT-User
Allow: /
User-agent: *
Allow: /
Sitemap: https://mysite.com/sitemap.xmlLimit : An open robots.txt isn’t enough if your host or firewall turns these robots away: that’s why the checker also makes a real request.
Yes. GPTBot is used to train models; ChatGPT’s search goes through OAI-SearchBot and ChatGPT-User. Blocking GPTBot is a choice, with no effect on citations in search.
Letting GPTBot in can help the model know your brand for answers without a search. Blocking it keeps your texts out of training. Both choices make sense.
Check it yourself : block training without blocking search
User-agent: GPTBot
Disallow: /
User-agent: OAI-SearchBot
Allow: /
User-agent: ChatGPT-User
Allow: /Not always. Most AI robots don’t run JavaScript: if your text only shows up once scripts load, they see an almost empty page.
Sites rendered on the server (WordPress, Shopify, Next.js with server rendering, static sites) usually don’t have this problem. Single-page apps that load content in the browser do.
Check it yourself : what a robot gets without JavaScript
In your browser, view the page source (Ctrl+U, or Cmd+Option+U on Mac) and search for a sentence from your text. If it isn’t there, AI robots most likely don’t see it. The online test runs the same check for you.
It isn’t needed today. llms.txt is a proposed format to summarize a site for language models, but robots barely read it and Google says it doesn’t use it.
Publish it if it’s easy for you: it costs almost nothing. Just don’t expect any measurable effect. On Shopify, the store already serves one on its own.
TadaSEO’s audit reports a missing llms.txt as information, never as a problem.
They help engines understand who you are and what your page says. A well-written FAQ mostly provides short answers that are easy to quote as they are.
Structured data (JSON-LD) describes your business, products or questions and answers in a machine-readable format. It doesn’t replace clear text: what gets cited is the visible content.
A good FAQ answer starts with the answer itself, in one or two sentences, then adds detail.
Limit : Google now shows FAQ rich results only for a few official sites; structured data still helps understanding, not a guaranteed display.
Because ChatGPT’s search, like Copilot, relies heavily on Bing’s index. A site well indexed by Bing has a head start.
Open a Bing Webmaster Tools account: it can import your Search Console property in a few clicks. Submit your sitemap, and turn on IndexNow if your CMS supports it, to report each new page right away.
There’s no guaranteed timeframe. Fixing robot access can work within days; new content usually takes several weeks to be indexed, read and picked up.
The most reliable approach is to measure the same questions every week, in the same conditions, and watch the trend rather than a single result.
Limit : No one controls assistants’ answers. A site can be cited one week and not the next without changing anything.
No. Neither TadaSEO nor anyone else can guarantee a citation in ChatGPT, or a first position on Google.
What you can do: remove what blocks you (robots, indexing, unreadable content), publish clear and sourced answers, be mentioned beyond your own site, then measure the effect week after week. Be wary of any offer that promises a citation.
How TadaSEO tracks ChatGPT, why a result can differ from what you see, and how to check it yourself.
Every week, TadaSEO asks ChatGPT the questions your customers ask (10 per site), then records whether your site is cited, at which rank, and which competitors and sources appear.
Questions are asked to a neutral ChatGPT, with no history or memory, from your site’s location. You can also test a question on demand for 2 credits.
Weekly tracking is included in the subscription, with no credits.
Limit : A measurement is one answer at one moment. The trend over several weeks is more reliable than a single result.
Because your ChatGPT is personalized: it knows your history, your memory and your exact location. Your customers’ ChatGPT doesn’t know you.
TadaSEO asks a neutral ChatGPT to see what a new customer would see. Answers also change from one time to the next and by location: two different results don’t necessarily contradict each other.
Check it yourself : see what a customer sees
Open ChatGPT in a temporary chat (no memory), or in a private browser window without signing in. Ask the question like a customer, for example “Which plumber should I call in Austin for an urgent leak?”, then look at the sources cited under the answer. Try two or three times: the answer can change.
ChatGPT is tracked on every plan. Perplexity, Gemini and Claude will come next.
With Google Analytics or Umami connected, you also see the visits sent by ChatGPT, Perplexity, Gemini, Copilot and Claude, whatever the assistant.
In your analytics tool, look at referring sites: chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com or claude.ai. Connected to TadaSEO, Google Analytics or Umami groups them by assistant for you.
Limit : Part of the visits from an assistant arrive with no referring site (mobile app, copied link) and are counted as direct. The figure shown is a minimum.
The basics that really matter, and how to check them.
Search Console is Google’s free tool that shows your real clicks, impressions, queries and positions. Connected to TadaSEO, it lets Tada base its advice on your actual performance.
Access is read-only: TadaSEO changes nothing in your Google account. You can remove it at any time, from TadaSEO or from your Google account permissions.
Limit : Search Console data arrives 2 to 3 days late: that’s how Google publishes it.
In Search Console, paste the page address into the “URL inspection” bar: Google tells you whether it’s indexed, and if not, why.
A page set to noindex, blocked by robots.txt, missing from the sitemap or never linked from your other pages is unlikely to be indexed. With Search Console connected, TadaSEO checks the indexing of your main pages every week and lets you know.
Check it yourself : a quick check in Google
Type site:mysite.com/my-page into Google. If the page shows up, it’s indexed. The opposite isn’t proof: only URL inspection is reliable.
About 60 characters for the title and 120 to 155 for the meta description. Beyond that, Google cuts the display.
The title starts with what the searcher is looking for, with the brand last if there’s room. The description promises the page’s concrete answer. A clearer title can win clicks without changing position.
Limit : Google sometimes rewrites the title or description it shows. Good text lowers that risk without removing it.
Yes, as one factor among others. Google measures the mobile experience with Core Web Vitals: an LCP under 2.5 seconds and a CLS under 0.1 are considered good.
LCP measures how long the page’s largest element takes to show. CLS measures layout shifts while loading. The free audit records both on mobile.
Limit : A fast site doesn’t rank on its own: content relevance weighs far more than a few tenths of a second.
It lists the pages you want indexed, so Google and Bing find them quickly. Declare it in Search Console and in your robots.txt.
Check it yourself : the line to add at the end of your robots.txt
Sitemap: https://mysite.com/sitemap.xmlThe free audit, credits, plans, connections and your data.
In about 30 seconds, the audit reads your robots.txt, crawls up to 30 pages, measures mobile speed and your domain’s authority, and checks whether AI robots can read you. No account, no credit card.
The report ranks issues by priority: tags, error pages, broken links, structured data, AI robot access. It also prepares 3 of your customers’ questions, tested on ChatGPT once you create your account.
The report stays available for 12 months, at an address that can’t be guessed and isn’t indexed by search engines.
Limit : One free audit per site every 30 days, and at most 3 a day from the same connection. The robot respects your robots.txt.
1 credit = 1 message to Tada or 1 Reddit search. 2 credits = 1 question tested on ChatGPT on demand. 10 credits = 1 written piece of content. Automatic tracking (Google, ChatGPT, Reddit) costs no credits.
Monthly credits reset at each renewal. Pack credits stay valid for 12 months, even if you stop your subscription. Tada uses monthly credits first, then packs, oldest first.
The cost of any action over one credit is shown first, and nothing goes through without your consent. 5 credits are offered at signup, once per site.
Starter tracks 1 site on Google and ChatGPT. Pro tracks 5, adds automatic publishing, Tada Continuous, MCP and API, and lets you add sites for €1.99 a month each.
You can upgrade to Pro at any time, prorated. Without a subscription, a credit pack gives access to Tada and writing, but not to automatic tracking.
There’s no trial: the free audit and the 5 credits offered at signup take its place. After that, the 14-day guarantee refunds your first payment if TadaSEO isn’t right for you.
The guarantee applies to your workspace’s first subscription, once, if you used at most 25% of the month’s credits (50 on Starter, 125 on Pro). The details are in the terms of sale.
Yes, from Starter and optional. Google Analytics connects in 1 click; Umami (Cloud or self-hosted version 3.4 or later) with an API key.
TadaSEO reads your visits, their sources, your landing pages and the visits sent by AI assistants, read-only. One tool per site. The audience analysis (1 credit) turns them into a plain summary and ideas added to your actions.
Limit : The figures are your tool’s: if it doesn’t record a visit (ad blocker, tracking refused), TadaSEO doesn’t see it either.
Yes. WordPress and Shopify connect with a step-by-step guide, to publish and fix in one click. For other sites (Webflow, Next.js, AI-generated sites), you export to Markdown or HTML, or your coding agent applies the actions through the MCP server.
The audit, Google tracking and ChatGPT tracking work with every public site, whatever the CMS.
Never on Starter: everything goes to drafts. On Pro, only if you choose it: a scheduled publication, or Tada Continuous in “publish for me” mode.
You can also keep approving each piece of content, and switch modes whenever you like.
On Pro, Tada writes and publishes articles and pages at your pace: you choose how many per week or month, the days, the time and the topics.
Tada finds topics in your data (Search Console, questions where you aren’t cited, pages to refresh), or follows your themes. Planned content shows in your calendar ahead of time. Each piece costs 10 credits; when credits run out, Tada pauses and resumes on its own.
Limit : Without a connected CMS, content stays in TadaSEO: you publish it yourself.
Yes, on Pro, through the MCP server: your assistant reads your actions, report, audience and content, then applies fixes in your code. A REST API offers the same operations.
Reads are free. Paid operations (asking Tada, ChatGPT test, writing, audience analysis) require an explicit cost confirmation: nothing is charged without it.
It’s only used for your tracking and analyses, hosted in the European Union. Google and analytics access is read-only, and you can remove it at any time.
Messages sent to Tada and your site’s public content are processed by our artificial intelligence subprocessor to produce answers. They aren’t used to train models. When you disconnect an analytics tool, synced data is deleted.
In one click from your account, with no commitment. Your subscription stays active until the end of the paid period.
Pack credits stay valid for 12 months after purchase, even without a subscription.
Write to us and a person will answer. Or run the free audit: in 30 seconds, you’ll know where your site stands.