AI SEO With Semrush: Build a Practical 90-Day Plan
You don’t need another list of 100 blog ideas. You need to know which page to improve next, which topic is worth writing about, and whether any of that work is actually paying off.
Connect Semrush to Claude or ChatGPT, and you can turn real search data into a focused 90-day SEO plan. I used this workflow on SolopreneurTools.io to find existing ranking opportunities, uncover content gaps, and compare visibility in Google and AI search. Here’s how to build a plan you can actually execute.
Connect Semrush to your AI assistant
You need Semrush and either Claude or ChatGPT. Semrush supplies the search data; your assistant helps organize it and explain what to work on.
My partner link offers a 14-day Semrush One Starter trial for new users. That gives you time to try the workflow before deciding whether the ongoing subscription fits your business. Check the checkout terms before signing up.
Set up the connection
Find Semrush in your assistant’s app or connector directory, choose Connect, sign in to Semrush, and review the requested access. Semrush’s current setup guide covers both Claude and ChatGPT. You can then tag @Semrush in a chat and ask for data.
This connection uses MCP, which simply lets the assistant request information from Semrush. Available data and usage limits depend on your plan. MCP queries consume API units, so start with a focused request rather than asking for every keyword worldwide.
Give it your business context first
Tell the assistant your domain, target country, audience, main offer, and available weekly time. A keyword with lots of searches is useful only if it attracts people your business can help.
My website is [DOMAIN]. I help [AUDIENCE] with [PROBLEM] and want visitors to [BUSINESS GOAL]. Use the [COUNTRY] database. I can spend [HOURS] per week on SEO. Use current Semrush data, label estimates, and flag anything you cannot verify.
You can also get the free SEO and AI Search Prompt Guide. The prompts below adapt the same workflow for a practical first pass.
1. Improve pages that are already close to winning
Start with keywords where your site ranks around positions 5 to 30. You already have a page Google considers relevant. Improving that page may be a better use of your next few hours than starting from zero.
Use Semrush to find my best existing SEO opportunities for [DOMAIN] in [COUNTRY]. Focus on keywords ranking 5–30. Return the top 20 with ranking URL, current position, search volume, keyword difficulty, and intent. Group keywords by page and rank the pages by business relevance and realistic improvement potential.
In my analysis, Claude grouped opportunities across eight existing articles and suggested changes for each page. That grouping matters: five related keywords might belong to one stronger article, not five new posts.
Turn a keyword into an actual page edit
Choose one relevant page, inspect the search results, and ask what the reader still needs. Maybe the introduction takes too long to answer the question, the tutorial has outdated steps, or the comparison misses a buying decision.
Review [PAGE URL] against the intent of [KEYWORD]. Suggest specific changes to the title, opening answer, headings, examples, and internal links. Separate evidence from assumptions. Preserve existing affiliate links and useful original content.
Keep the URL unless there’s a clear reason to change it. Your goal is a more useful page, not a rewrite for the sake of freshness.
2. Find low-competition keywords worth your time
Once you’ve identified existing opportunities, look for relevant topics you haven’t covered well. My starting filter is keyword difficulty below 10, combined with meaningful search demand.
Use Semrush to find new keyword opportunities for [DOMAIN] in [COUNTRY]. Start with keyword difficulty below 10 and relevant search volume. Exclude topics unrelated to [OFFER]. Return keyword, volume, difficulty, intent, and suggested page type. Check whether an existing page could cover each topic before recommending a new article.
Semrush’s keyword difficulty metric estimates competition on a 0–100 scale. A low score helps you narrow the list; it doesn’t guarantee a ranking.
Use relevance to break the tie
If two topics look achievable, choose the one closest to a problem your customer is trying to solve. A smaller, specific search can be more useful to your business than a broad topic with far more volume.
Check the current results before committing. Are readers looking for a tutorial, a calculator, a product page, or a comparison? Build the format that solves the problem. If the first list is weak, give the assistant feedback and ask it to search within a narrower topic.
3. Find the gaps your search competitors reveal
Your search competitors aren’t always companies selling the same thing. They may be publishers or specialist blogs answering the questions your customers ask.
Use Semrush to identify the most relevant organic search competitors for [DOMAIN] in [COUNTRY]. Explain the keyword overlap. Find valuable keywords they rank for that I’m missing or ranking poorly for. Prioritize relevant, lower-difficulty opportunities and show the competing URLs, volume, difficulty, and recommended next action.
Review the competitor list before accepting the gap analysis. A huge website with a few overlapping terms may be less useful than a smaller site serving the same audience.
Then ask what you can add. Your own walkthrough, screenshots, examples, or a clearer explanation can make the article useful. Copying another site’s headings won’t give readers a reason to choose yours.
Combine this list with your existing-page opportunities. If a gap belongs in an article you already have, add it to that refresh instead of creating competing pages.
4. Check AI visibility separately from Google rankings
Traditional SEO asks where your pages rank. AI visibility asks whether your brand is mentioned or your pages are cited in AI-generated answers. Those are related questions, but they need separate measurements.
For my demonstration, I used browser control to inspect Semrush’s AI Visibility reports because that report wasn’t available through the MCP connection I was using. If your connector cannot retrieve it, open the report in Semrush and use a supported browser assistant, or review it yourself.
Analyze the Semrush AI Visibility report for [DOMAIN]. Show the available platforms, report date, brand mentions, cited pages, and relevant topics. Compare the most relevant competitors. Identify three opportunities and explain the evidence for each. Flag unavailable data instead of filling in gaps.
Look at which pages and sources appear for the questions you want to answer. A useful next step might be improving an explanation on your own site or identifying a relevant publication where your expertise could contribute.
Treat the report as a measured view of the queries and platforms it covers, not a complete record of every AI answer. A mention is also different from a click or a customer. Keep those outcomes separate in your plan.
For the bigger picture, read GEO vs SEO: why your website still matters in AI search.
5. Turn the research into a realistic 90-day SEO plan
At this point, you have four inputs: existing rankings, new keyword ideas, competitor gaps, and AI visibility. The next step is deciding what fits your time.
Using the findings above, build a prioritized 90-day SEO and AI visibility plan for [DOMAIN]. Assume [HOURS] hours per week. For each task, show the page or topic, supporting evidence, exact action, estimated effort, dependency, and success measure. Prioritize business relevance and achievable gains. Include what to skip for now. Do not invent missing metrics.
Days 1–7: Fix the foundations and record a baseline
Run a site audit, check important pages for crawl or indexing problems, and record current rankings before changing anything. Prioritize issues that affect useful pages. A higher site-health score alone is not the business outcome.
I used an AI browser assistant to help correct technical issues in Squarespace. You can also ask for step-by-step instructions. For changes such as merging pages or redirects, review the proposed URLs and keep a record of what changed.
Days 8–30: Refresh the strongest existing pages
Work through a short list of pages with relevant rankings and clear improvements. Update outdated details, answer the main question earlier, and add useful supporting examples. Record the date and purpose of each edit.
Days 31–60: Create content that fills a real gap
Choose a manageable number of new topics from the combined research. Build each around a distinct reader need. Connect it to relevant existing content through natural internal links.
Days 61–90: Distribute, measure, and refine
Share useful resources with your audience, pursue relevant collaborations, and review what improved. Carry successful ideas into the next cycle. Drop work that attracts the wrong audience or depends on resources you don’t have.
This schedule is a starting structure. Let your actual audit findings and available time change the order. A good plan should tell you what to postpone as clearly as what to do next.
Track progress without reacting to every ranking change
Save a baseline and review it every two to four weeks. That’s a useful reporting rhythm, not a promise that search engines will reward your changes within that time.
Compare [DOMAIN] against my baseline from [DATE], using the same country, device, keyword set, and report source. Show gains and losses by keyword and page, plus overall visibility. Separately summarize AI mentions and citations where comparable historical data exists. Relate changes to my edit log, flag uncertainty, and recommend the next three actions.
In my early check, one ranking moved from 37 to 27 after recent changes. That was encouraging, but a short-term movement doesn’t prove which edit caused it or whether it will hold.
Watch outcomes beyond position: relevant visits, resource downloads, enquiries, or sales. If your workflow also uses Instagram to deliver resources and collect leads, my Claude and Blotato DM automation guide covers that next step.
Use the data to choose the next useful improvement. Constantly rewriting a page before you can evaluate the previous change makes the results harder to understand.
Final takeaway
The hardest part of SEO is often deciding what deserves your attention. Semrush gives you evidence; Claude or ChatGPT helps turn it into a practical sequence of work.
Start with pages already showing promise, validate the gaps, and build a plan that fits the time you actually have. Better decisions and consistent follow-through matter more than publishing another pile of content.
Ready to build your own 90-day plan? Try Semrush One Starter free for 14 days with my partner offer for new users, and grab the free prompt guide to follow along.
Kevin Fernando
The accompanying video was sponsored by Semrush. Some links are affiliate links. I may earn a commission at no extra cost to you.
