GEO
Claude optimization
Claude searches Anthropic's own index and cites fewer sources per answer than any other engine. That makes it the hardest one to enter and the most valuable to hold. The work is narrow: allow the Anthropic crawlers, be present on the review sites and curated lists it prefers, and publish pages that need no interpretation.
30 minutes. No pitch deck. You leave with a prioritised plan.
Claude is the quietest of the answer engines and the strictest. It searches Anthropic's own index, picks a small number of sources, and writes a careful answer that leans on them. There is no long list of footnotes to hide in. This page covers the part of our GEO services aimed at that engine.
Why is Claude the hardest engine to enter?
Because there are fewer places to stand.
3.6
An engine that shows a dozen sources gives an adequate page a reasonable chance. An engine that shows three or four does not. Every citation is a choice against the alternatives, so the page has to be the clearest available statement of the thing being asked, not merely a competent one.
The upside is symmetrical. Being one of three sources is worth far more than being one of twenty, and once a page becomes the thing Claude reaches for, it tends to stay there while it remains accurate.
Where does Claude get its sources?
From an index Anthropic builds itself.
Claude does not read Google's results or Bing's. It searches its own index and, when a question points at a specific page, fetches that page live. So ranking well in a search engine is not a route in. The route in is being in the index, being readable, and being the kind of source it prefers.
Anthropic split its crawlers into separate agents early in 2026, which matters because a blanket rule written before that split can block the wrong one.
| What matters | How Claude works |
|---|---|
| Where the answer comes from | Anthropic's own index, with a live fetch when a question points at a specific page |
| Crawlers to allow | Claude-SearchBot for the index, Claude-User for live fetches, ClaudeBot for training |
| What it tends to cite | Review platforms and curated comparison lists, and fewer sources per answer than any other engine |
No other engine reads Anthropic's index and no other crawler substitutes for these, so a rule written for one engine does nothing here. The full six-engine comparison sits on the GEO services page.
What does Claude cite when someone asks who to hire?
Other people's judgements about you.
For vendor and shortlist questions, Claude leans on review platforms and curated comparison lists rather than vendor marketing. Community threads, which carry real weight in other engines for general topics, barely register here. That shapes the plan more than anything on your own site does.
So the work splits in two. Off your site: complete and correct profiles on the review platforms your buyers actually use, real reviews collected through a process you can show anyone, and inclusion in lists written by people who are not you. On your site: pages that make it easy for a list author to describe you accurately, which means published specifics rather than adjectives.
What do we change on your pages?
We remove the interpretation work.
Claude tends to quote statements it does not have to reconstruct. So every target page opens with a two or three sentence answer under the heading. Headings become the questions a buyer asks: what it includes, what it costs, how long it usually takes, what happens if it does not work. Facts get a named source and a date. Adjectives that cannot be checked come out.
Then we check what the crawler actually receives. A page whose text is painted by a script after load can arrive nearly empty, and Claude will simply cite someone else without telling you. Moving that text into the server-rendered HTML is ordinary technical SEO work and it is usually the biggest single change we make.
What does the work look like?
Baseline
Run your agreed prompt set in Claude and record whether you were named, whether you were linked, and which three or four sources it used instead of you.
Open access
Live-fetch as Claude-SearchBot and Claude-User, fix any rule or firewall preset that blocks them, then fetch again to prove it. Decide on ClaudeBot separately.
Earn the third-party pages
Complete the review profiles, run an honest review process, and pitch inclusion in the curated lists the baseline showed Claude reading.
Rewrite what it should quote
Answer at the top, buyer questions as headings, one table, sourced numbers, and the text present in the raw HTML.
Re-run and report
The same prompts every week, logged as named, linked or absent, with the competing sources named so the next task picks itself.
How do we measure it?
With a fixed prompt set and an honest scoreboard.
Because Claude cites so few sources, small changes look dramatic and single runs mean little. We run the same 20 to 40 questions weekly and read the trend rather than the week. The report logs three states per question, lists the sources that appeared instead of you, and says which of them we are working on.
Claude sends less referral traffic than Perplexity, so we do not judge it on sessions. We judge it on whether the shortlist a buyer is handed includes you.
The neighbouring engine pages are ChatGPT optimization and Perplexity optimization. If you want to see how we run this for a team in one of our regional markets, our market page walks through the same process.
Who should prioritise Claude?
Companies selling something considered, to people who read before they buy.
Claude is used heavily for work research: drafting, analysis, comparing options before a recommendation goes to somebody senior. If your buyer is the person preparing a shortlist for a decision-maker, the answer they get shapes the document that decision-maker reads. That is a small audience with unusual weight.
It matters less for impulse categories and for businesses whose customers arrive through a sales team or a marketplace rather than through research. In those cases we still open the crawlers, because the cost is an afternoon, and we spend the writing and outreach budget on the engines your buyers actually use. Which engines get funded is a decision from the audit, not a default.
What does not help here?
Publishing an llms.txt file. No answer engine uses it to decide citations, Google confirmed in June 2026 that it has no effect, and a study of 137,000 domains found 97 percent of published files were never requested once.
Adding FAQ markup to chase a rich result. Google retired FAQPage rich results in May 2026. We keep the markup where it mirrors visible text, and we keep the visible questions, because those were always the part doing the work.
Volume. Publishing thirty thin pages to increase your chances does the opposite here, because a small citation budget goes to the clearest source, not the most frequent one.
How do we start?
Book a free 30-minute consultation. We check whether the Anthropic crawlers can reach you, whether your key text exists in the raw HTML, and what your markup says, then name the first three fixes.
Plans are flat monthly retainers with starting prices on the pricing page. They run month to month, and you can stop at the end of any month.
Questions we get asked
Where does Claude get its sources?
From Anthropic's own index, built by its own crawler, plus live fetches when a question points at a specific page. It does not read Google's or Bing's results, so ranking in either one does not put you there.
Which crawlers do we need to allow?
Claude-SearchBot builds the search index, Claude-User fetches a page live when a user's question calls for it, and ClaudeBot collects training data. The first two decide whether you can be cited at all.
Why is Claude harder to get into than the other engines?
Because it cites the fewest sources per answer. Where another engine lists a dozen pages and gives you room to be one of them, Claude usually names three or four. There is less space, so the bar for each one is higher.
What does Claude tend to cite for vendor questions?
Review platforms and curated comparison lists written by someone other than the vendor. Community threads barely appear for these questions. That makes third-party presence the main lever rather than page volume.
Does blocking ClaudeBot hurt our search visibility?
Not directly. ClaudeBot is the training crawler. Claude-SearchBot and Claude-User are the ones that affect citations. Deciding about training is a separate business decision, and we will lay out both sides rather than push you either way.
How long does it take?
Crawler fixes change what Claude can read within days. Page work usually surfaces over four to eight weeks. Getting onto the review platforms and lists it prefers usually takes two to four months, because it depends on other people's editorial cycles.
Can you guarantee Claude will name us?
No. There is no way to control the output, and with only a few citations per answer the competition for each slot is real. We commit to the work and report weekly on the prompt set we agreed.
How is it priced?
As part of one flat monthly retainer, with starting prices on the pricing page. Most of the access and page work serves every engine at once, so covering several engines in one plan costs less than running them separately.
Thirty minutes to see where your site is invisible.
Book a free consultation with Ibrahem and Jeremy. We look at your site on search, AI answers and conversion while we talk, and you leave with a prioritised plan.
30 minutes. No pitch deck. You leave with a prioritised plan.