← BlogHow-to

How to get recommended by Claude and Gemini, step by step

How to get recommended by Claude and Gemini, step by step: what each assistant reads, the settings that decide it, and how to check the answer changed.

To get recommended by Claude and Gemini, publish a page that answers each question your customers ask, plainly and near the top, and check that your site settings let both assistants read it. Then ask each question again after the page goes live and record whether either assistant names you.

Claude and Gemini come from two different companies, Anthropic and Google. Each reads the web through its own maker's visitors, and each maker documents what your settings control. The page that earns a mention is the same for both. The settings you check are different, and that is where most of the work in this guide sits.

How Claude and Gemini decide who to name

Every AI assistant answers in one of two ways. It answers from the web, by reading pages at the moment someone asks. Or it answers from its memory, from what its model learned in training. Both Claude and Gemini can do both, and their makers say how.

Claude. Anthropic's help article on its crawlers, updated 7 April 2026 and read on 29 September 2026, names three visitors:

  • Claude-User reads pages while Claude answers someone: "When individuals ask questions to Claude, it may access websites using a Claude-User agent."
  • Claude-SearchBot "navigates the web to improve search result quality for users."
  • ClaudeBot collects web content that could contribute to training Anthropic's models.

Anthropic also states: "Anthropic's Bots respect 'do not crawl' signals by honoring industry standard directives in robots.txt." So your robots.txt file decides which of the three can read you.

Gemini. Google's crawler documentation, last updated 14 July 2026 and read on 29 September 2026, describes grounding as "providing content from the Google Search index to the model at prompt time to improve factuality and relevancy." A setting called Google-Extended controls whether your pages may be used for that grounding in Gemini Apps, and for training future Gemini models. The same page states that "Google-Extended does not impact a site's inclusion in Google Search nor is it used as a ranking signal in Google Search."

For Gemini, that means two things. Your page has to be in Google's search index, so Google has to be able to crawl it. And your settings must not opt it out through Google-Extended.

Whichever assistant is reading, the page it can use has the same three traits. It can be read. It answers the question plainly near the top. And its facts agree with the other places your business is described.

Do these in order. Each step removes a guess from the next one.

Step 1: List the questions your customers ask

Write down the questions a buyer types before they know your name, in their words. Say you run a payroll service for small restaurants. Your list includes category questions ("What is the best payroll service for a small restaurant?"), cost questions ("How much does restaurant payroll cost a month?"), comparison questions ("[your brand] vs [a rival]: which handles tips better?") and questions with your name in them ("Is [your brand] worth it?").

Aim for 10 to 20. That is enough to see a pattern and few enough to track by hand.

Step 2: Ask Claude and Gemini each question, and record both

Ask every question in Claude and in Gemini, the way a customer would. For each answer, write down whether you were named, who was named instead, which pages were linked and the date.

Keep the two assistants in separate columns. They can give different answers to the same question, because each has its own model and reads the web through its own maker's systems. A question you win on Claude can be a question you lose on Gemini, and the fix can differ.

Ask each question more than once, on different days. Answers change from one ask to the next. A pattern across several asks is worth acting on. A single answer is not.

Step 3: Check the settings for each assistant

Open your robots.txt file, or run Rankli's free check, and look at four names. To see what Claude says about your market today, the free Claude visibility checker asks it three buyer questions with web search on.

For Claude: Claude-User, Claude-SearchBot and ClaudeBot. Anthropic spells out what blocking each one does. Blocking Claude-User "prevents our system from retrieving your content in response to a user query, which may reduce your site's visibility for user-directed web search." Blocking Claude-SearchBot "may reduce your site's visibility and accuracy in user search results." Blocking ClaudeBot signals that your future pages "should be excluded from our AI model training datasets."

For Gemini: Google-Extended. Blocking it tells Google your pages may not be used to train Gemini or to ground answers in Gemini Apps. It does not change your place in Google Search. If you block it to stay out of training, know that the same setting covers grounding too.

Then check two things that apply to every assistant.

Words AI can see on your pages. Some sites build their text in the browser with JavaScript. A visitor that reads only what your server sends can miss those words. Your answer should be in the page as it is served.

Nothing that tells search engines to skip the page. For Gemini, the page has to be in Google's index. A tag or header that asks search engines not to index a page keeps it out.

Rankli's free check reads your site with no JavaScript, the way an AI visitor does, and reports how your settings treat Claude's visitors and Google-Extended by name. It asks no AI assistant anything, so it is free and needs no account.

Step 4: Publish an answer page for each question you are losing

Take each question from step 2 where someone else was named, on either assistant. Publish one page for it, built to be quoted:

  • A title that matches the question, in the customer's words.
  • The answer in the first one or two sentences.
  • Specifics a reader can check: prices, limits, places and dates.
  • The same facts your other pages state. Two versions of a price give an assistant two numbers to choose from.
  • A visible date.

One page per question lets each page answer its question in its first lines. The same page serves Claude and Gemini, so you write it once.

Step 5: Ask again after the page goes live

A week after the page goes live, ask the same question in both assistants again, a few times each. Compare with step 2. Then do it once more a week later.

Two checks after publishing, set against the answers from before, show whether the page changed anything on each assistant. Rankli runs the same schedule: 7 and 14 days after an article goes live, it asks the article's question again, 3 times on each assistant your plan tracks, and shows before and after on the article.

Where Claude and Gemini differ

The page is the same for both assistants. The machinery around it is not.

ClaudeGemini
Made byAnthropicGoogle
Where web answers come fromPages Claude-User and Claude-SearchBot can read, per AnthropicContent from the Google Search index, per Google
The setting to checkThose two, plus ClaudeBot for training, in robots.txtGoogle-Extended in robots.txt, and that Google can index the page
What the maker says about blockingBlocking Claude-SearchBot may reduce your visibility in user search resultsBlocking Google-Extended opts you out of Gemini grounding and training, not Google Search

Two practical points follow. A Gemini answer can only ground on a page Google has indexed, so a page Google cannot crawl is out of reach for Gemini's web answers. And a Claude answer depends on Anthropic's own visitors, so a rule that blocks them affects Claude even when Google reads you fine.

What not to do

Do not copy a robots.txt file without reading it. A file that blocks "all AI" can block Claude's search visitors and opt you out of Gemini's grounding in a few lines.

Do not assume one assistant speaks for the other. Record Claude and Gemini separately. A win on one tells you nothing certain about the other.

Do not judge on one answer. Ask each question several times, on different days, before you decide where you stand.

Do not publish two versions of your facts. Old pages with old prices stay readable. Update or remove them when you publish the new answer.

Do not change five things at once. Change the page, then wait for the re-ask. Changing the title, the answer and the settings in one week hides which change mattered.

How Rankli runs these steps for you

Rankli runs this loop every week, in four steps:

  1. Ask. Every week, Rankli asks ChatGPT, Claude and Gemini the questions your customers ask. Perplexity too, on Scale and Agency.
  2. See what you're losing. You see each question where AI recommends someone else, and who it picks instead.
  3. Publish the answers. Rankli writes an article aimed at each question you're losing and publishes it to WordPress, Webflow, Shopify or the other connected platforms.
  4. Re-ask and show what moved. 7 and 14 days after each article goes live, Rankli asks its questions again and shows you before and after.

Rankli asks each assistant with its web search turned on, then records who it names and which pages it cites. How Rankli tracks AI answers has the detail. The questions and articles in each plan are on the pricing page. For Perplexity, which links the pages it reads, see how to get recommended by Perplexity.

Rankli starts with a 7-day free trial. A card starts it, and starting it runs your first analysis: your questions put to ChatGPT, Claude and Gemini, with the ones you are losing listed. On day 8 it becomes your plan unless you cancel, which is one click in the app. Cancel during the trial and nothing is charged.

Questions

Does Claude read websites when it answers?

Yes. Anthropic says that when people ask Claude questions, it may access websites using its Claude-User agent. Rankli asks Claude with its web search turned on and records which pages it cites.

Does blocking Google-Extended hurt my Google rankings?

No. Google's documentation says Google-Extended does not affect inclusion in Google Search and is not a ranking signal there, though it does control whether Gemini may use your pages for training and grounding. Rankli's free check reports how your settings treat it.

Why do Claude and Gemini name different businesses?

Each assistant has its own model and reads the web through its own maker's systems, so the same question can get different answers. Rankli records each assistant's answer separately, so you see where you stand on each one.

Which Rankli plans track Claude and Gemini?

Every plan. Rankli asks ChatGPT, Claude and Gemini your customers' questions on every plan, and adds Perplexity on Scale and Agency.

How this post was made

The Rankli team wrote this post, and the product scored it. It went through the same 15 rules a customer’s draft goes through: the word-count range, the heading structure, the keyword placement, the fabricated-statistic detector and the rest. Every number below came back from that run on the words above it.

100SEO score15 of 15 rules passed
Written by
Rankli editorial
Target keyword
get recommended by claude and gemini
Words
1,835
Keyword density
0.98%
Reading grade
7.3
Internal links
5

Every rule, as it scored

hard blocks a draftadvisory lowers the score

  • Passed: word-counthard1835 words; the target range is 1500–3000
  • Passed: heading-hierarchyhardone H1, no skipped levels
  • Passed: keyword-in-titlehardtitle contains "get recommended by claude and gemini"
  • Passed: keyword-in-openinghardthe first 100 words must mention "get recommended by claude and gemini": searchers and crawlers both decide there
  • Passed: keyword-in-h2hardat least one H2 must contain "get recommended by claude and gemini"
  • Passed: keyword-densityharddensity 0.98% of words; above 2.5% reads as stuffing: vary the phrasing
  • Passed: banned-phraseshardno banned phrases
  • Passed: no-fabricated-specificshardno unsourced statistics, citations or quotes detected
  • Passed: internal-linkshard5 internal link(s); 2–5 contextual links to the customer's own pages
  • Passed: subject-coveragehardmentions "answer", "rankli" outside the keyword
  • Passed: section-substancehardevery section carries at least 25 words
  • Passed: title-deliversadvisorythe title makes no specific promise to check
  • Passed: meta-description-lengthadvisory151 characters; aim for 150–160 so it fills the snippet without truncation
  • Passed: title-lengthadvisory57 characters; titles over 60 are truncated in results
  • Passed: readabilityadvisoryreads at grade level 7.3; above 12, sentences are likely too long for a general audience

The rules are the ones in src/content/seo.ts, unchanged for us: hard rules block a draft from ever reaching a customer’s site, advisory rules only lower the score. Scored and committed as a markdown file, so the commit is the receipt. Run the same fifteen checks on a draft of your own in the SEO article checker, or see what it costs.

Rankli writes posts like this for your site.

Same fifteen checks. 7 days free, then your plan. Cancel before day 8 and nothing is charged.

Start free trial

Keep reading

All posts →