Competitor and SERP Research With Claude

#01What can Claude actually see on the web?

Three things matter here. First, web search: Claude can search the web during a chat and cites its sources, and on Free and paid plans web search is on by default (Anthropic). Second, web fetch: when you give Claude a URL, it can open and read that page. Third, Research: on paid plans, Research runs "multiple searches that build on each other" and decides what to look into next based on what it finds (Anthropic).

What Claude doesn't have is a search engine's results page as you see it. When you ask "what ranks for X", Claude runs its own searches through its search provider and reports what it finds. That's useful for understanding which pages and angles dominate a topic, but it isn't a rank tracker. For positions, volumes and click data, use your SEO platform and Google Search Console, then bring the exports to Claude for analysis.

Our view: treat Claude as the analyst who reads everything, and your SEO tools as the instruments that measure. Asking Claude for search volumes is like asking a colleague to guess them; they might be close, but you wouldn't put the number in a report.

#02How do you build a competitor profile?

Start with one competitor and a clear template, so every profile is comparable. Ask Claude to open each competitor's own pages first, then reviews and press coverage.

Build a competitor profile for [Competitor A, url].
Open their homepage, pricing page, product pages and 3 recent
blog posts. Then search for independent reviews and news from the
last 12 months.

Return:
- Positioning in one sentence, in their words (quote + URL)
- Target customer and main use cases
- Pricing model and published prices (with URL and date seen)
- 5 claims they make most often
- What reviewers praise and criticise most (with sources)
- Recent changes: launches, pricing changes, hires, funding
- Gaps: questions their content doesn't answer

Cite every point. Mark anything you couldn't verify.

Run the same prompt for each competitor in separate chats inside a project called "Competitor research", and save each profile to the project as a file. That keeps the knowledge in one place and stops one long chat from running out of context.

When the profiles are done, ask for a comparison in a fresh chat: "Using the profiles in this project, build a comparison table of positioning, target customer, pricing model and top claims, then tell me where our positioning is most and least differentiated." This is often the single most useful output of the whole exercise.

#03How do you analyse a search results page?

Choose the queries that matter to your business, then ask Claude to study the pages that appear for them. You can either let Claude search, or paste in the top URLs from your own search or SEO tool, which is more reliable because you control the list.

Here are the top 10 URLs for the query "[query]":
[list]

Open each one and tell me:
1. The format (guide, list, comparison, tool, product page, forum)
2. The main angle and the question it answers first
3. Its H2 headings
4. Original evidence it uses (data, examples, expert quotes)
5. What's missing or weak

Then summarise: what does a page need to compete here, and what
angle is nobody covering well?

This gives you a content gap analysis grounded in real pages. The "what's missing" column is where your opportunity usually sits. If every result is a generic list, a page with original data or a worked example has a real chance.

Some sites block automated fetching, so Claude may not be able to open every URL. Ask it to tell you which ones it couldn't read rather than guessing their content.

#04When should you use Research instead?

Use Research when the question is broad, needs many sources, or you don't yet know what to look for. Good examples for marketers include sizing a new market segment, understanding how buyers in a new region evaluate tools like yours, or mapping every player in an adjacent category.

Use research. I need an overview of [category] tools for
[audience] in [region]. Identify the main players, how each
positions itself, typical pricing models and the buying criteria
that reviews and analyst pages mention most. Prioritise sources
from the last 12 months and cite each finding.

Anthropic notes that Research can also use your connected tools, such as Google Drive or Gmail, when started on the web or desktop, and that research started in the mobile app only searches the web (Anthropic). It also says research sessions use your limits faster, so save it for questions that deserve it.

A Research report is a starting point. Read the sources it cites for any finding you plan to act on, and be especially careful with market-size figures, which are often recycled between low-quality pages.

#05How do you track competitors every week?

Scheduled tasks let you describe a check once and have Claude run it on a schedule. Anthropic says scheduled tasks are available on all paid plans, run remotely even when your computer is asleep, and list "recurring research" on topics, competitors or industry news as a common use (Anthropic).

Every Monday at 8:00, check these competitors: [names + URLs].
Look at their pricing pages, product update or changelog pages,
blog, and news coverage from the past 7 days.
Report only changes since last week: new launches, price changes,
new positioning, major content. Link every item. If nothing
changed, say so in one line.

The "report only changes" instruction keeps the digest short enough that you'll actually read it. Add your competitor profiles to the files the task can reach, so Claude has a baseline to compare against.

#06How do you check that the research is right?

Three habits cover most risks:

1. Open the key sources. For anything that changes a decision, click through. Pricing in particular changes often and may differ by region.
2. Check the dates. Ask Claude to put a date next to each finding. A competitor's "new" feature might be two years old.
3. Separate fact from interpretation. Ask Claude to label which points are directly stated on a page and which are its inference. "Competitor B targets enterprise" might be an inference from one case study.

Also be careful about what you share. Don't paste confidential sales notes into a chat on a plan where your organisation hasn't approved that use. Check your data controls and your company's policy first.

#07How should you name competitors in shared outputs?

Within your team, use real names. In anything public, follow your company's rules on comparative claims, and make sure any comparison is accurate, current and fair. AI engines and readers both notice when a comparison page only ever favours the publisher, and inaccurate claims about competitors carry legal risk in many markets. When in doubt, ask your legal team before publishing.

#08From MentionX data: how competitors show up in AI answersOriginal data

Method. Read-only aggregate queries over MentionX scans from 22 August to 29 October 2026 (UTC): 2,217 engine runs (2,154 completed) across 18 anonymised projects and six engines.

Competitor research now includes AI answers too. MentionX's cross-engine aggregates show how often different engines mention brands in the prompts we track:

Our view: add a short AI-visibility section to each competitor profile: which engines mention them and for which prompts. Our guide to measuring AI share of voice explains how.

Limits. A convenience sample of mostly B2B software projects, run through engine APIs, which can behave differently from the consumer apps. Perplexity, Copilot and DeepSeek figures come from a single day (26 October 2026), so treat them as a snapshot, not a trend.

#09Worked example: Ledgerline maps the approvals market

Ledgerline is a fictional accounts-payable tool. Figures are illustrative.

Ledgerline's product marketer builds profiles of four rivals, Competitors A to D, in a "Competitor research" project. The comparison chat shows that all four lead with speed, while none explains multi-entity approvals well.

She then pastes the top ten URLs for "invoice approval workflow" from her SEO tool. Claude finds seven generic list articles, two product pages and one forum thread. Nobody offers a worked example with real approval limits. That becomes Ledgerline's next guide.

A Monday scheduled task now reports competitor changes. In its third week it flags that Competitor C has removed its free plan, which she confirms on the pricing page before telling sales. Ledgerline and these details are fictional.

#10What should you try before the next lesson?

Profile your closest competitor with the template above, then run the SERP prompt on one query you care about. Compare the gaps Claude finds with your content plan. If you have a paid plan, set up the weekly scheduled task and see whether the first digest is useful.

Next: mining customer reviews for messaging. Or return to the Claude for Marketers track.

#11How do you turn research into positioning decisions?

Research only pays off when it changes what you say or do. Once you have profiles and a SERP analysis, run a short positioning session with Claude in a fresh chat in the same project.

Ask Claude to list every claim your competitors make and group them into themes. Then add your own main claims. The themes everyone shares, such as "fast", "easy" or "secure", are table stakes: you need to support them, but they won't differentiate you. The themes only you can credibly claim are your positioning candidates. Ask Claude to challenge each candidate: what evidence supports it, which competitor could easily copy it, and what a sceptical buyer would say.

Then test the strongest candidate against real customer language. Lesson 8 shows how to mine reviews and support tickets for the exact words customers use. A positioning claim that matches how customers already describe their problem is far more persuasive than one invented in a workshop.

Finally, write down what you've decided and why, with links to the evidence. Positioning documents age quickly, and a record of the reasoning makes it much easier to update when a competitor changes direction.

#12How do you research a new market or region?

The same tools apply, with two adjustments. First, ask Claude to search in the local language as well as English, and to tell you which sources are local. Buyers in many markets rely on local review sites, publications and communities that English searches miss. Second, set the location context in your prompt, such as "for buyers in Germany", because search results and competitors differ by country.

Ask for a short list of the local sources Claude found most useful, then check a few yourself. If you have colleagues or partners in that market, share the findings with them before acting. A short review from someone who knows the market will catch assumptions that no amount of web research can.

#13What are the limits to keep in mind?

Claude's research is only as good as what's publicly available and readable. Private communities, gated reports and pages that block automated access won't appear. Sales conversations, win-loss interviews and analyst briefings often contain the most valuable competitive insight, and they need people. Use Claude to process those notes once you have them, but don't expect it to find them on the open web.

Keep one more limit in mind: Claude's answers vary between runs, because its searches and the pages it reads can differ each time. For anything you'll repeat, such as a monthly market check, use the same prompt and the same list of URLs so the results are comparable from one month to the next, and note the date each check was run.

#14Common questions

Can Claude show me what ranks on Google? Not directly. Claude runs its own web searches. For accurate rankings, export the top results from your SEO tool and ask Claude to analyse those pages.

Is Claude Research available on the free plan? No. Anthropic says Research is available on paid plans (Pro, Max, Ultra, Team and Enterprise) and needs web search turned on.

Can Claude monitor competitors automatically? Yes, on paid plans. Scheduled tasks can run a competitor check on a set cadence, even when your computer is asleep.