A website is never finished and needs refreshing.

How we decide which pages to update, which to leave alone, and what a review of a B2B client’s site showed us.

A B2B website is usually judged on launch day. Every page is written, the design is approved, and the project closes. Buyers don’t visit on launch day, though. They arrive months and years later with new questions, and more and more often after reading an AI-generated answer above the search results.

At ShikhaShikz we write content with a long shelf life, and a page that still ranks well and serves its readers is left alone. But shelf life varies. Figures go out of date, pages written for yesterday’s questions turn thin, and some never answered the buyer’s real question in the first place. A content review shows which pages need work. That is why a website is never finished. 

This post covers what Google says about content for AI search, how we review a page, and what we found when we compared a B2B client’s site with its direct competitors.

A website is a working asset, not a finished project.

What the AEO and GEO advice actually says

There is a lot of noise about answer engine optimisation and generative engine optimisation. 

Google treats optimising for AI search as a matter of SEO best practices. It says you don’t need llms.txt, special markup, or content chopped into small pieces for AI. It asks for first-hand, non-commodity content and gives one test: would your visitors find it satisfying? It also warns against third-party tools that promise ranking success. Google says nothing specific about updating pages, so we won’t claim it endorses refreshes.

Bing is more direct. It says regular updates help AI systems reference the current version of your content, and it recommends clear headings, tables, FAQ sections, and claims backed by data and sources. In a separate post, Bing explains why this matters: stale facts can produce wrong answers, and its AI can decline to use a source whose support is missing, stale or conflicting. 

The two engines don’t give identical advice, but they agree on the basics: useful, accurate, well-organised content. Where content has aged, keeping it accurate is the part a team can control. 

What other teams are doing

Most are already using AI in their content work. In a survey of about 300 marketers by Ahrefs, 93% of respondents used AI to create or help create content. The most common uses weren’t writing from scratch. They were brainstorming, outlining and improving existing content. Nearly nine in ten said they check what AI produces by hand.

We read that as a sensible pattern. Use AI to speed up analysis and drafting, and keep a person responsible for accuracy. This is a self-reported survey from an SEO software vendor, so it shows what teams say they do, not what works.

How we review a page

We run the same six checks on every piece of content:

1. Search Console data for 28 days, three months and one year, to see whether the page is gaining, holding or fading. Google’s Generative AI performance report in Search Console also shows how often links to the site appear in AI Overviews and AI Mode, which is worth checking alongside.
2. Google Analytics data for the same three windows to see what visitors do once they arrive.
3. Keywords. Check the keywords already in the page against their search volume, then add the relevant ones that Search Console shows the page appearing for.
4. Sources. Replace old reports and figures with the latest.
5. The introduction. Whether it still fits or needs a fresh opening.
6. Related questions. Add a section only when there is a real gap. Adding sections for their own sake makes a page longer, not better.

The result of these checks is often to change nothing. A page that is ranking well and still accurate stays as it is. Only pages that are fading, outdated or thin are edited. 

What we found on our B2B client’s site

We reviewed their website against a set of established checks and compared it with their direct competitors, and it revealed a recurring pattern, one we often see across B2B websites. 

• The website pages were thinner than those of their direct competitors. By ‘thin’ we don’t mean word count alone, because Google says there is no ideal length. The pages said less than a buyer needs.
• The blog was uneven. Structure changed from post to post, cover images were inconsistent, and the language dialect wasn’t consistent.
• Few pages answered a question directly. A reader had to dig for the answer, and so would any system trying to lift one.

Why that matters

Thin pages cost them in three ways.

First, they weaken conversion. A buyer gets teaser-level information and is then asked to contact sales, so the page does little to move them closer to a decision.

Second, they make the site a weaker candidate for AI-generated answers. Our assessment is that deeper pages with clear structure are more likely to be cited. Google’s guidance to organise pages into clear sections and offer first-hand content supports part of that, but Google does not promise it, so we treat it as an insight and not a rule.

Third, they leave topic territory to competitors. The competitors cover process, materials, engineering logic, testing and supply-chain concerns in fuller detail, so they own more of the same problem space the client wants to be known for.

What we proposed and where it sits in our work

This review isn’t a standalone task for us. It runs across three services we already deliver, and each recommendation maps to one of them.

• SEO. Review every page against Search Console and Analytics data, and give each page a direct answer near the top.
• Content strategy. Restructure the blog into one consistent format, expand the website pages to cover what a buyer needs, and fix the cover images for the blog posts and language dialect so the site reads as one voice.
• Demand generation. Give buyers enough on the page to arrive at sales better informed, and review key pages on a schedule so they stay accurate.

These are the outcomes we are working toward: stronger support for buyers before they reach sales, a better chance of appearing in AI answers, and a wider share of the topics their competitors currently cover. In our follow-up posts, we will cover how content enhancement worked for them. 

A website is never finished. It is reviewed on a schedule, and the review is what keeps it useful.