AI visibility audit10 min read
ByKarol PiaseckiFounder TextRank.ai
Published
Reading time10 min read
Updated

Competitor Content Analysis Example: From Source to Page Edit

Follow one buyer prompt through the model's answer, the sources it surfaced and your page, until it becomes a specific content change your team can ship.

Illustrative example, available without signing up. The company, answer and edit below demonstrate the workflow. This is not a customer audit report or evidence of increased citations, traffic or sales.

Competitor content analysis is easiest to assess through one question and one actionable task. It supports a broader AI visibility audit by providing a specific edit plan. So let's take one buyer prompt and follow it all the way: through the model's answer, the sources it surfaced, and into a page change a writer could ship this week.

One prompt, one buying decision

Say a company sells contract-management software to mid-sized procurement teams. A buyer types:

Which contract-management tool is suitable for a mid-sized company that needs fast implementation and a CRM integration?

There is a lot packed into that sentence. The buyer already knows the category, so a definition won't help. What they have is three criteria: company size, implementation effort and CRM integration. A page that answers those has a real shot at the shortlist. A page about "streamlining your contract workflows" probably doesn't.

First, we record the model, the date, the full answer and the surfaced sources. We also note who can verify the facts later. The integration owner knows which CRMs are supported and on which plan. The implementation lead knows how long setup really takes and what the customer needs to bring. That step feels like admin, but it saves the most time down the line.

In TextRank, you can audit one page or a group of pages, up to three of your own URLs per prompt (buyer question). For an offer-level question, a product page, the pricing page and the documentation usually make a good set. That is exactly where a buyer would look for answers to their three criteria.

What do the other sources have that your page doesn't?

The sources a model surfaces tell you a lot about which information turned out to be useful. Assume they explain implementation stages, name the supported CRM systems and say which plan includes each integration. The company's product page covers the same ground with two phrases: "fast onboarding" and "seamless integrations."

Put side by side, the gap is obvious:

Buyer criterionUseful detail in the surfaced sourceCurrent wording on the product page
implementationstages, required roles and conditions"fast onboarding"
CRM supportnamed systems and sync method"seamless integrations"
fit for a mid-sized teamsuitability and limitationsfeature list for everyone

The word "fast" is doing heroic work here. Fast for a five-person team importing 200 contracts? Fast with a dedicated implementation manager? Fast compared with what? Until the page draws a boundary around the claim, the buyer can't weigh it against anyone else's.

Here's the good news: the company usually knows those answers. It just hasn't written them down on the page. Closing this gap doesn't need a new feature, only a clearer description of what already works.

It's also worth being honest about what we haven't shown. We haven't proved the page was left out because of those two phrases. We have found a credible content gap: other sources answer the buyer's criteria more fully. That's reason enough to improve the page.

How AI Content Optimization turns a gap into a change plan

Spotting a gap is diagnosis. AI Content Optimization takes the next step: it shows which passages to keep, what to clarify and which evidence to gather so the page answers the selected prompt more completely. Every recommendation comes from comparing your page with the model's answer and competitor sources, so the writer always knows where it came from.

In our example, the brief covers three kinds of material.

Missing sections and topics. The product page needs a dedicated implementation and integrations section that opens with a direct answer. Not a longer intro, but specifics: who does what, how long it takes, on which plan.

Evidence to source. Before the "fast implementation" claim gets expanded, someone needs to collect the facts: the stages, a typical timeline for a team of that size, what the customer has to provide, and the list of supported CRMs. The brief also names who should sign those facts off.

Sections to improve. "Fast onboarding" and "seamless integrations" are flagged right where they appear. The team decides whether to rewrite, qualify, back up or remove them.

The brief also suggests a format when it makes information easier to use. For this prompt, a short "plan → supported CRMs → sync method" table and two or three implementation FAQs would work well. It surfaces relevant keywords and entities too, such as the CRM names, but as part of the brief rather than decoration sprinkled through the copy.

The same pattern works in any industry. The feature page walks through an ecommerce platform that described itself as providing "everything you need to grow your online business." A competitor, in the same spot, stated launch time, named integrations and how often stock levels sync. The direction for the fix was similar: costs, launch time, integrations, sales channels and plan differences, plus a dedicated comparison page with clear selection criteria.

From vague advice to a task someone can finish

A weak audit often ends with this:

Add implementation keywords and expand the integration section.

It sounds actionable until a writer tries to use it. Which systems? Which claims are safe? What counts as done?

A brief worth assigning is narrower and answers those questions up front:

In "Implementation and integrations," replace the general claims with the verified onboarding stages, the roles required on the customer side and the supported CRM systems. Link each technical claim to documentation. Product approves the integration facts; Implementation approves the timeline and conditions.

That task has a location, a reason, two approvers and a clear finish line. You can assign it on Monday and review it on Friday. If the company can't verify the timeline, "fast" should go, or be qualified with only the conditions the team can stand behind. That's an improvement too: the page stops promising something the buyer can't check.

You keep the last word

The brief speeds things up, but it doesn't replace what your team knows. Only you know what implementation really looks like for customers, which integrations are polished and which are still on the roadmap, and how your buyers actually talk. The workflow leaves room for that: you can correct the premise of a suggested answer, add facts no source has, and make the page sound like your company rather than a generic category description.

Before publishing, you have a list of the facts, examples and approvals still outstanding. Nothing goes live unchecked, and nobody has to dig through threads to find out who was supposed to confirm what.

The edit is only half the job

After publication, save the exact passages that changed. Run the same prompt against the same models under comparable conditions, then compare the new answers and sources with your first observation.

One retest may show a different set of sources. It may show an answer that looks much the same, because models and their sources change at their own pace. Neither result undoes the improvement. The page now answers a real buying question with facts anyone can check, and that helps your sales team, your implementation team and buyers arriving from search, too.

Record two separate observations: what improved on the page, and what changed (or didn't) in the model's next answer. Keeping them apart is less exciting than declaring a win after one test, but it gives you a solid footing for the next decision. And the next prompt on your list can follow the same path, only faster.

From one prompt to the whole page

This example covered one prompt and one section. In practice, it pays to work through several questions from different stages of the decision: a head-to-head with a named competitor, pricing, security, implementation. Each can point to a different missing detail. Together they give you a prioritized change list your team can actually finish, instead of one big "let's rewrite the website."

Want to see this on your own pages? Start with the question your buyers ask most often and run it through AI Content Optimization. For the complete method, read how to run an AI visibility audit, and for picking the right questions, see how to choose buyer prompts for an AI audit.

An excerpt from an illustrative report

Every field below demonstrates the format rather than recording a model run.

FieldIllustrative content
QuestionWhich contract-management tool fits a mid-sized company using a CRM?
ModelChatGPT — selected for illustration, no actual run
Date and model versionNot applicable to the illustration; record the date and available model identifier in a real pilot
Market and languageUnited States, English — demonstration context
ScopeProduct page, pricing and documentation; up to three of your URLs
AI answer — illustration“Compare implementation stages, required roles and integration scope with your current CRM.”
Sources — illustrationA competitor's implementation guide and integration documentation; in a real report, check the exact URLs and content availability
Current passage“Fast onboarding and seamless integrations.”
GapMissing stages, requirements, CRM names and synchronization conditions
PriorityStart with “Implementation and integrations” because it addresses this question's criteria
RecommendationAdd stages, roles and a supported CRM table linked to documentation
Facts to confirmSetup time, supported systems, plans, limits and support
Approval and publicationAn expert verifies facts, a writer prepares the section and the page owner approves and publishes

Draft for the team to complete:

Implementation includes [verified stages]. Your team provides [roles and data]. The [CRM name] integration is available on [plan] and synchronizes [scope] under the conditions in [documentation]. Typical setup takes [verified range and conditions].

This is a draft. Every missing fact needs confirmation before publication. This example does not show an approved customer edit or an observed result after publication.

What does it cost to assess a plan like this?

A full ChatGPT audit costs 7 credits. Gemini / Google AI Overviews also costs 7 credits and shares one TextRank audit result, counted once. This includes the answer, available sources, comparison with your content, gaps and an edit plan.

A 50-credit pack costs $25 net or PLN 95 net. Using 7 credits from that pack costs $3.50 or PLN 13.30 net; you purchase the full pack. One question in ChatGPT and Gemini / AI Overviews uses 14 credits. One observation before an edit and one after uses 28. This is a small pilot of plan quality and workflow, not evidence of effectiveness. More repetitions require more credits. See the scenarios and calculator.

How to record pilot results

Before starting, define questions, URLs, market, language, models, repetitions and assessment criteria. Keep full answers and sources where available. The shared Gemini / Google AI Overviews result is one observation, not two independent results.

RecordBefore the editAfter the edit
Period and runsEnter dates and NEnter dates and N
Brand mentionAnswers with a mention / NAnswers with a mention / N
Edited URL citationAnswers citing the URL / NAnswers citing the URL / N
Use of the added informationRecord the passage or absenceRecord the passage or absence
Time to prepare a planMeasure for the defined scopeMeasure for the same scope
Page changeSave the original textRecord publication, date and verified facts
Other eventsModel changes, campaigns, website updatesRecord the same types of events

Report no improvement and deterioration too. Distinguish a brand mention, a source link and an actual product recommendation. Treat tool scores, content quality, citations, traffic and sales as separate outcomes. One observation before and after cannot separate an edit's effect from answer variability.

A real case study requires source records, customer permission, the edits, observation counts and limitations. Statements or a demonstration alone do not justify a growth claim.

Frequently asked questions

What does this example show?

The full path: from a buyer's question, through the model's answer and surfaced sources, to a specific writing task and a retest after publication.

Is one prompt enough for a complete brand audit?

One prompt is enough to fix a specific passage and test the method. A fuller picture needs questions from several stages of the buyer's decision, from category comparisons to pricing and implementation.

What exactly does AI Content Optimization give me?

A brief tied to the prompt, the page and the observed gap: missing sections, weak passages flagged in place, a suggested answer format, and a list of the facts and approvals still needed before publication.

Will the model start recommending my company once I make the changes?

No one can honestly guarantee that. What you get is a page that answers a real buying question with verified facts, plus a record that lets you compare later model answers with your earlier observation.

See how to hand over the edit plan to a writer: passage, reasoning, facts to approve and acceptance condition.