Most sites are invisible to search. Now they are invisible to AI too.

Pages marked non-indexable, plus any noindex tag or robots.txt rule blocking a page you want ranked

Zero pages blocked that you intend to rank.

Remove the blocking directive, then request indexing in Search Console.

Indexed pages against sitemap pages, and sitemap processing status

90% or more of submitted pages indexed, sitemap processed within the last 30 days.

Resubmit the sitemap, fix blockers found in the previous step, and improve internal links to unindexed pages.

Pages returning a server error or a not-found, and the internal links and images still pointing at them

No server errors at all. Under 1% of your internal links pointing at pages that no longer exist.

Repoint or remove the broken links. Redirect any missing page that still has other sites linking to it.

Redirects that pass through an intermediate URL before landing, and internal links pointing at a redirect rather than the final page

No chain longer than one hop.

Point every redirect straight at its final destination.

Canonical tags pointing at a redirect, a broken page, another canonical, or looping back on themselves

Every canonical tag points, in one step, at a page that loads normally.

Point each canonical at the live final URL.

Core Web Vitals report, mobile and desktop

75% or more of visits pass on every metric.

Use the lab audit in the next step to find causes, fix them, then confirm here. Field data lags about 28 days.

The Lighthouse performance score, how long the main content takes to appear, and which scripts or stylesheets are holding up the render

Performance 90 or above on the page types that carry the most traffic.

Work the Lighthouse opportunities in order of estimated saving, largest first.

Pages with no internal links pointing at them, and how many clicks each important page sits from the homepage

No page left without internal links pointing at it. Every important page within three clicks of the homepage.

Add a link to each stranded page from a relevant parent or hub page.

Whether each template carries structured data, whether it validates, and whether the type matches what the page actually is

Every page type carries markup, no validation errors, and the type matches what the page actually is.

Fix errors first, then check the type is right. The crawl proves the markup exists. The Schema Markup Validator proves it is valid. Only the Rich Results Test proves Google will actually use it.

Pages not on HTTPS, secure pages linking out to insecure ones, missing mobile viewport settings, and unusually heavy pages

Everything on HTTPS, no secure page linking to an insecure one, and the mobile viewport tag set on every page type.

Force HTTPS and repoint internal links at the secure URL.

Reciprocal tags, valid language and region codes, a self-referencing tag on each page

Each language version lists every other one, including itself, and a default version is set for everyone else.

Rebuild the set so each page lists every alternative version including its own.

Every question tagged with where it came from, separating what you observed from your own buyers and data from what was inferred from search tools

At least two different sources, one of them your own evidence. Every question carries where it came from.

Add your own sources. Search Console queries are free and already yours. Interview transcripts and support tickets usually exist and are rarely read.

A list of seed terms expanded into long-tail variations, each with monthly search volume and a difficulty score, filtered to what your domain can realistically reach

Every target term carries a volume and a difficulty, and the list is filtered to difficulty your current authority can win.

Start from your seed terms, expand them with a keyword tool, and pull volume and difficulty for each. Then cut everything above the difficulty your domain rank supports and work up from the bottom.

Pages with no title, titles too short or too long, and the same title used on more than one page

Every page unique, under 60 characters, primary term near the front.

Rewrite duplicates and overlong titles, starting with the pages that already get impressions.

Pages with no description, and the same description used on more than one page

Every page unique, between 120 and 160 characters.

Write descriptions for the pages with high impressions and low click-through rate first.

Pages with no H1, with more than one, or with headings that skip levels

Exactly one H1 per page, describing the page rather than the brand.

Add the missing H1 and demote any duplicates.

Pages carrying very little text for their size, low word counts, and poor readability scores

No page you want indexed sitting below the word count that page type needs.

Merge, expand or remove. A page worth keeping is worth finishing.

Named authors with real credentials and a biography page, a named reviewer on sensitive pages, publication and updated dates, outbound citations to authoritative sources, an editorial policy, and first-hand experience visible in the writing

Every page that matters names its author, shows a published and updated date, cites its sources, and on sensitive subjects names a qualified reviewer.

Name the author and give them a real biography page. On anything about health, money, safety or law, name a qualified reviewer and publish an editorial policy. Cite an authoritative source for every claim, and let first-hand experience show rather than claiming expertise.

Subjects that every page on the first results page covers and yours does not

Your page covers every subject the ranking pages share, in your own words and with your own examples.

Add the missing subjects to the page you already have rather than starting a new one, and keep your own numbers and examples in rather than rewriting to match everyone else.

The angle each ranking page takes, the format it uses, and whether yours repeats one of them or offers something they all miss

Your page takes a defensible angle the ranking pages do not, in the format the results reward, with your own data, examples or experience behind it.

Read the first ten results and write down the angle each one takes. Pick the angle nobody has covered, or pick the most common one and out-execute it with first-hand detail they cannot copy.

Ranked keywords filtered to positions 11 to 20, sorted by volume

Every striking-distance term with real volume has an owner and a next action.

Refresh the ranking page, add internal links to it, and match the intent the results are actually rewarding.

Competitor ranked keywords minus yours, filtered by relevance and difficulty

A prioritised gap list, scored by volume against difficulty.

Work the list in order of volume divided by difficulty, not alphabetically and not by what is most fun to write.

Whether each topic has one pillar page with supporting pages linking into it, and whether any cluster is half-built

Every topic you claim has a pillar page and enough supporting pages to cover the question properly, all internally linked, with one clear owner page per cluster.

Pick the topic you most want to own, build the pillar page, then fill the cluster around it before starting a second topic. Half-built clusters win nothing.

Page types in the top ten, plus the Labs intent classification

Your page type matches the dominant type in the top ten.

Change the format to match, or pick a different term. Reading the live results beats trusting the label.

More than one URL receiving impressions for the same query, or swapping positions over time

One clear owner page for each query.

Merge them, make them clearly different, or point the weaker page's canonical tag at the stronger one.

Images with no alt text

Every meaningful image described. Decorative images explicitly empty.

Write alt text describing the image, not stuffed with keywords.

Referring domains, domain rank, new and lost over time

Referring domains trending up over 90 days, losses explained.

Work the link gap and broken link steps below. Publish something worth citing.

Domain intersection across two or more competitors, minus your own referring domains

A live target list, prioritised by domain strength and relevance.

Domains linking to two or more competitors and not to you are the warmest targets on the list.

The spam score of every referring domain, pulled in bulk across the whole profile

No meaningful share of links from high spam score domains.

Disavow only clear cases, which means telling Google to ignore those links. Most sites never need to, and over-disavowing does real damage.

Competitor pages returning 404 that still hold referring domains

A list of dead pages with live links, matched to your equivalent page.

Offer your page as the replacement, and only where it genuinely covers the same ground.

Brand mentions across the web, with sentiment, filtered to those with no link

Checked at least monthly, with a standing outreach routine.

Ask. A short note to the author converts a good share of these.

Impressions and clicks for queries containing your brand name, tracked over time, and the monthly search volume of the brand name itself

Branded impressions rising quarter on quarter, and your own site holding the first result for your own name.

Branded search grows from everything off the site: press, partnerships, social, community, paid awareness, word of mouth. It cannot be optimised on the page. Pick one channel and run it long enough to move.

Anchor distribution across referring domains

Mostly your brand name and natural phrasing, with your target search terms used as link text only a small minority of the time.

Earn more brand-anchored links rather than trying to remove the others.

Name, address and phone consistency across listings and review platforms

Identical details everywhere, profile claimed and complete.

Correct the outliers, starting with Google Business Profile.

Live position for each tracked keyword, weekly

A tracked set that is checked weekly and reviewed monthly.

Check weekly, not quarterly. Fifty keywords weekly for a year costs about $5.

Analytics installed on every page, Search Console verified and linked to analytics, conversions defined and firing, internal and team traffic excluded, and the organic channel attributing correctly

Conversions are defined and recorded, internal traffic is excluded, Search Console and analytics are linked, and organic sessions reconcile with Search Console clicks within a sensible margin.

Fix tracking before acting on any number in this section. Define the conversions that matter, exclude your own devices, link Search Console to analytics, and check that a test conversion actually records.

Clicks, impressions, click-through rate and average position over 28 days against the prior period

Clicks trending up over 90 days.

Work the striking distance and click-through rate steps before creating anything new.

Queries in the top ten with click-through rate well below the expected curve for their position

No high-impression query sitting far below its expected rate.

Rewrite the title and description for that query's intent. Check whether a results feature is absorbing the click.

Which features appear on the results page for each query: AI overview, featured snippet, people also ask, ads, local pack, video

You hold the features that exist for your priority terms, or you know why you do not.

Format the answer to win the snippet, or accept a lower ceiling and reweight the term.

Organic sessions, engagement and conversions by landing page in Analytics, then the same visits traced through to deals in your CRM

Organic conversion rate at or above your site average, and you can say which deals started with an organic visit.

If traffic rises and conversions do not, the problem is intent match or the page itself, not ranking.

Traffic against rankings against links over the same window, plus what the page looked like before the drop

Any drop has a named, evidenced cause.

Stable rankings with falling clicks is a click-through or results-layout problem. Falling rankings site-wide is algorithmic or technical. A drop confined to one page group is a content problem.

Your ranking positions month by month, and what the results pages looked like at the time

Trend understood and dated against known update windows.

Line the dates up against algorithm update windows before concluding anything.

Competitor visibility over the same window as your own drop

You can name who took the position and what they changed.

Read the page that replaced yours. Usually it answers the question better or more directly.

View source on each important page, or turn JavaScript off in the browser, and compare the text actually in the HTML against what the page shows. To check the whole site at once, crawl it twice in Screaming Frog, once with JavaScript rendering on and once off, and compare the word counts.

The text sitting in the HTML is within 20% of what the page shows once it has loaded, on every important page.

Render the text on the server so it is in the HTML before JavaScript runs. No amount of AI optimisation helps a page that returns nothing without JavaScript.

Share of questions from the bank where the brand is mentioned, by platform

Mentioned in a meaningful share of category questions, rising quarter on quarter.

Fix readability first, then proof and comparison content. Present this as a snapshot, because answers vary between runs.

Competitor mention counts and cited domains across the question set

You know the top cited domains for your category and have a route into them.

Target the repeatedly cited sources. Presence in them is a faster route to citation than site changes alone.

Whether an AI overview appears for the query at all, and whether you are cited inside it

Cited in overviews on your priority terms, or you know which terms trigger one.

Overviews draw on the sources ranking beneath them. Winning the underlying results is still the route in.

Organisation and person markup carrying one consistent identifier, links out to your official profiles, and one clear description repeated the same way everywhere

One official description of the business, used the same way on the site, in the markup and on your social profiles.

Use the same description everywhere, and make sure the markup points at the same identifier on every page.

The wording ChatGPT, Claude, Perplexity and Gemini each return about the brand, scored for accuracy. Record all four, because they disagree and they draw on different sources

Descriptions match your own positioning, with no outdated pricing, features or claims.

Find the source of the wrong claim and correct it there. Old press and outdated third-party listings are the usual culprits.

Named authors, dates, original figures, named customers, methodology, direct quotes

Every page that matters carries at least one specific fact an AI assistant could quote and attribute.

Add original data and named sources. One real number beats a page of claims.

Coverage of alternatives, versus and best-for queries from the question bank

An honest page for every serious alternative, including where you are not the right choice.

Write the comparisons. Naming where a competitor is the better fit increases the odds of being cited, not the opposite.

llms.txt present, and robots.txt rules for GPTBot and OAI-SearchBot (ChatGPT), ClaudeBot and anthropic-ai (Claude), PerplexityBot, and Google-Extended (Gemini and AI overviews)

A deliberate decision, documented. Either allowed and guided, or blocked on purpose.

Check robots.txt before anything else in this section. Blocking the crawlers makes every other step here pointless.