--- title: "Off-Page AI Search Visibility: A 90-Day Plan" url: https://adtech.lol/articles/7d8f0523-7ea9-449c-91ec-aebc375958ec source_url: https://searchengineland.com/audit-off-page-ai-search-visibility-90-day-plan-494866/ source_site: searchengineland author: "Kristina Frunze" published: 2023-10-01T00:00:00+00:00 tags: ["SEO", "AI", "B2B SaaS", "Visibility Audit", "Digital Marketing"] publisher: AdTech Radar --- # Off-Page AI Search Visibility: A 90-Day Plan > A structured approach to auditing AI search visibility for B2B SaaS brands to improve market performance and visibility. # Introduction Running an SEO and AI search optimization agency for B2B SaaS brands, I needed an AI search audit that would uncover actionable insights, identify visibility gaps, and give my team a clear direction — without creating unnecessary busywork. After plenty of experimentation, I’ve developed a standard operating procedure (SOP) that we now use internally. It focuses on off-page AI search optimization, since much of your brand’s AI visibility depends on third-party websites. ## Take a snapshot of where your brand stands today As a prerequisite for this process, you need to create a project dedicated to your brand using your AI visibility tool of choice (there’s no shortage of those nowadays). At our agency, we use Peec.ai (we’re not affiliated with the company — it’s just a tool we’ve stuck with, and I like its functionality). Each tool might have a different setup, but most popular tools offer similar reports or variations of them. ### Come up with prompts First, you’ll want to nail down your prompts. My advice is to stick to bottom-of-funnel (BoFu) prompts that people search for during the decision-making stage — these will most likely influence your pipeline. The way we come up with BoFu prompts is simple — start with the target BoFu keywords you’d absolutely want to show up for. Since we work with SaaS brands, many of these keywords have modifiers such as “software,” “solution,” and “tool.” For example, the keyword “construction scheduling software” could become “What’s the best construction scheduling software for general contractors?” if your ideal customer profile (ICP) is general contractors. ![](https://searchengineland.com/wp-content/seloads/2026/10/Best-construction-scheduling-software.png) You’ll want to have 20-50 prompts for your brand, depending on the size of your business, the industries you serve, and the offerings or use cases you provide for these ICPs. Remember, there’s no “perfect” prompt. You just want to get close enough to the questions your ICP would typically ask LLMs. Another great way to enrich your prompt pool is to listen to sales conversations and speak with your customers to find out what really matters to them when they’re looking for a solution like yours. ### Set up a project in your tool Once you have your prompts, make sure to select your key competitors within the tool so you’re tracking your AI performance against your competition — not in a vacuum. This helps you understand the competitive landscape when making strategic decisions. Ideally, you’ll want to let the project run for at least a couple of weeks so you have enough data to start making observations and drawing conclusions. ## Run an AI visibility audit We use Claude MCP extensively to automate mundane and time-consuming tasks, including the audit itself. We’ve built an internal skill based on the SOP that does the heavy lifting for us, but it still requires someone to steer the process and make decisions, so we’ve built evaluations (evals) along the way. ### Step 1: Pull visibility and share of voice for all nonbranded prompts What Claude does for us: - Pulls visibility and share of voice for all prompts, then splits the data by engine. - Diagnoses the visibility patterns using this table: | | | | | --- | --- | --- | | **Brand is weak in** | **Likely cause** | **Who fixes it** | | Google AI Overviews only | Rankings — AI Overviews lean on what already ranks | Agency (SEO) | | ChatGPT only | Weak entity presence and mentions across the web | Client (listings) and agency | | Perplexity only | Citation sourcing and freshness (it favors recent content, directories, and Reddit) | Agency (outreach and community) | | One topic, every engine | Content gap | Agency (content) | This step establishes an initial snapshot of the brand’s performance across three major AI search engines and helps plan further actions. ### Step 2: Pull top citations for your prompts AI engines build answers from a fairly small set of sources they keep returning to. Knowing which domains and pages those are tells you exactly where influence lies in this category. What Claude does: - Pulls the top 40 domains based on how often engines retrieve them, followed by the top individual URLs. - Pulls the search queries the engines fan out to behind the scenes. These show retrieval intent and feed directly into content planning. - Manually reclassifies Peec’s source categories, since its labels are sometimes unreliable (G2 has been classified as user-generated content). Categories used: Owned, Competitor-owned, Directory/review, Editorial/media, Community/UGC, Ecosystem, and Analyst. - Adds Domain Rating and traffic data from Ahrefs for context. - Flags any competitor whose own site gets retrieved in a large share of chats. That usually means they publish comparison and alternatives content that models treat as a category reference, and it often explains most of the leader’s advantage. - Samples 20-30 top-cited URLs and records their format, publication or update date, and whether the client appears. The key output of this step is a map of the sources that shape AI answers in this category, including the best-performing formats and where the client’s content falls short. ### Step 3: Identify source gap URLs (the outreach list) Producing listicles on your own website is half the battle for LLMs. To have the best chance of showing up in LLMs, your brand needs to be named in as many third-party category listicles as possible. This step helps us build a list of the most-cited listicles and comparison pages. What Claude does: - Filters Peec for URLs classified as Listicle, Comparison, Alternative, Profile, or Discussion, where at least two competitors appear and the client is absent, sorted by retrieval volume. - Assigns tiers: - Tier 1 = 25+ retrievals. - Tier 2 = 5-24. - Tier 3 = fewer than five. - Marks any page owned by a vendor or competitor as “DO NOT PITCH.” A competitor’s own “X vs. Y” article will never add the client. The output is an LLM Outreach file that we’ll use as the first line of action to secure mentions in those publications. ### Step 4: Checking the sentiment and objections When generating answers, LLMs repeat certain objections about brands (“expensive,” “limited integrations,” “steep learning curve,” etc.). With Peec, we can see the sources of these claims and correct potentially inaccurate information where applicable — tracing an objection back to an old pricing page is gold. What Claude does: - Pulls the objections AI engines raise about the client, along with brand attributes and how the client ranks against competitors on them. - Pulls the source URLs behind each reported objection. - Assigns each objection a status: - **Outdated:** Fix the source page. - **Competitor framing:** Publish counter-content. - **Accurate:** Work on transparency and product feedback. - Before calling anything outdated, searches real customer reviews. If reviewers are still complaining about the issue, the objection reflects genuine feedback, even if the feature now exists. That’s framed as a lag between the product and the public record. The outcome is a table of objections, with the tracked source and a recommended action. ### Step 5: Check directories and entity consistency Directories, review sites, and social media profiles are among the sources LLMs pull from when doing research. Typically, missing, unclaimed, or outdated listings contribute to inaccurate claims about a brand. What Claude does: - Builds the list of properties from the data itself: directory, review, and community domains from Step 2, plus industry directories from the client’s knowledge base and standard platforms (LinkedIn, YouTube, and app stores where relevant). - Compares Ahrefs referring domains for the client against each competitor. A directory linking to competitors but not to the client is a likely gap. - Runs site: searches to find each listing and capture its URL. - Extracts fields where pages load properly: name as displayed, description, category, mobile app shown, pricing language, review count, rating, last review date, and claimed status. - Cross-checks listing descriptions against the client’s approved stats. A listing claiming a figure the client never approved gets its own row. - Assigns a status: PRESENT, NOT APPEARING (couldn’t find it, unconfirmed), DUPLICATE (confirmed), or CHECK. Fields Claude can’t confirm remain UNVERIFIED and are never guessed. As a result, Claude produces the “Directory & Entity Fix Tracker,” which becomes a working sheet we use to claim, create, and correct profiles and listings. ## Turn your findings into action over the next 90 days Over the next 90 days, the work focuses on three action streams that come directly from the audit: directory fixes, outreach, and objection fixes. Use the full report to prioritize these efforts and measure progress against the baseline established at the beginning of the process. ### Fix directory listings Working from the “Directory & Entity Fix Tracker,” we claim unclaimed profiles and correct descriptions, categories, and pricing language. It’s important to maintain consistent messaging across all directories to avoid confusing LLMs about your key offering. ### Get onto the pages AI cites Working from the “LLM Outreach” document, we pitch Tier 1 pages first, followed by Tier 2. Usually, this involves emailing the article’s author or an editorial contact at the publication. We do whatever it takes to get our clients mentioned there. ### Fix objections at their source Working from the objection table in the Off-Page AI Visibility Analysis report (Phase 5), we handle each objection based on its status: - **Outdated:** Update the source URL. Directory sources go into the Fix Tracker, our own pages go to on-site work, and third-party articles become update requests in outreach. - **Competitor framing:** Add counter-content to the content plan, using the format Step 2 showed is winning. ### Re-measure at day 90 Rerun the audit using the exact parameters established in Step 1 (same prompts, engines, competitor set, and window length). Compare nonbranded visibility by engine, share of voice, directory status counts, the headline objection, and AI referral traffic. ## Make AI visibility an ongoing priority AI search visibility isn’t static. The sources AI engines rely on, the competitors they recommend, and the objections they surface can change over time. Treat this audit as a repeatable process rather than a one-time exercise. Regularly monitoring these shifts will help you adapt your strategy as the AI search landscape evolves. Source: [https://searchengineland.com/audit-off-page-ai-search-visibility-90-day-plan-494866/](https://searchengineland.com/audit-off-page-ai-search-visibility-90-day-plan-494866/)