You have four hours a week for six weeks and no new software. Spend the 24 hours on pages already ranking at positions 4–15 for buyer questions that trigger Google AI Overviews, not on another tool or a batch of new posts.

That is the organic version of working a striking-distance keyword: Google already finds the page relevant, but the page has not earned the result you want. I would rather make that page easy to crawl and easy to quote than start five posts from zero. Existing pages first. Answer-first rewrites next. No software purchase in week one.

The constraint: 24 hours, six weeks, zero new tools

Four hours a week is not a motivational limit. It is what a busy owner or solo marketer can spare after running the business. It decides what makes the cut. Each step below costs four hours and buys one thing the next step needs: a question list, a page shortlist, crawlability, extractable answers, proof, then a read on what changed.

The free stack covers this scope: Search Console for queries and URL Inspection, PageSpeed Insights for Core Web Vitals, and Rich Results Test for schema. If you cannot yet check ten questions and a handful of pages by hand, you do not have a software problem.

Who should skip this plan

Skip it if you need leads this week, have no pages ranking at all, or need a site rebuild before Google can read your pages. I used to tell clients any visibility work helped any business. I was wrong. Paid search can buy inquiries now; this plan gives existing pages a better shot at citations over six weeks. If Search Console shows almost no impressions for buyer questions, or Google rarely shows an Overview for what you sell, put the four hours into reviews, referrals, and one tight service page instead.

Use this plan when Google already sees your business but does not cite it yet.

1. Week 1: Log 10 buyer questions and their Overviews

Cost: 4 hours. Buys: a target list worth spending the other 20 hours on. Write down ten questions buyers ask before they pay you, in their words. Not a keyword export. Think “how much does garage door spring replacement cost” or “how long does probate take in Texas.”

Search each question yourself in a clean browser while logged out. Put three observations in a sheet:

  1. Does an AI Overview appear?
  2. Which sites does it cite?
  3. Does it cite your site?

Add the URL on your site that should answer the question. That gives you a starting record to revisit in Week 6. In a 300-query check of small-business searches, 280 showed an Overview. The same check found YouTube cited in 57% and Reddit in 20%, while independent blogs were near 1%. Do not treat those figures as a forecast for your ten searches. Look at your actual results.

The familiar ranking-guide template is easy to find: question heading, direct answer, short paragraphs, lists. It will not tell you which buyer questions trigger Overviews in your market. That is why the manual log stays. Mark questions without an Overview as non-triggering rather than pretending every query belongs in this plan. When one shows an Overview citing competitors, note what their pages answer that yours does not.

What gets cut: volume. Ten questions, not fifty. Pages covering fan-out sub-queries were more likely to be cited than pages covering only the main query, but depth takes time. You cannot cover every follow-up this week.

Leave Week 1 with ten rows: trigger status, cited sites, and your matching URL.

2. Week 2: Find the pages one push away

Cost: 4 hours. Buys: a ranked work list, not a wish list. Open Search Console, go to Performance > Search results, and review the queries with impressions. Compare them with your ten questions, then check the average positions of the matching pages. Keep the buyer questions where an existing page sits around positions 4–15. Do not build the plan around a page nobody has found yet.

This is the same triage instinct I used in PPC: work the relevant terms with room to improve before polishing everything in the account. One agency reports more organic leads after improving its striking-distance pages with depth and internal links. Your job here is narrower. Sort the pages into three piles:

  • Pile A: A page ranks around positions 4–15 for a buyer question that triggers an Overview, but your site is not cited. Fix these first.
  • Pile B: A page ranks in that range, but the question shows no Overview. Keep it for later.
  • Pile C: Everything else. Leave it alone during these six weeks.

Position does not entitle a page to a citation. In an Ahrefs re-run covering 863,000 keywords, only 38% of cited pages ranked in the top ten for the same query. A page at position 8 is not a near-certain winner. It is an existing asset you can improve without starting cold. If you want to see the approach applied to an existing ranking page, read that before editing Pile A.

What gets cut: the tempting pages you like but buyers do not find for these questions. Three to five Pile A pages are enough to inspect. You will rewrite only two. Do not turn a four-hour shortlist into a ten-page wish list.

3. Week 3: Remove the crawl and snippet blockers

Cost: 4 hours. Buys: pages Google can read and potentially use. Google does not require special AI markup or an llms.txt file for a page to appear as a supporting link in AI Overviews. The page needs to be indexed and eligible for a snippet, as this optimization rundown explains. Start with the ordinary blockers: robots.txt restrictions, noindex, nosnippet or data-nosnippet instructions, and important text that depends on JavaScript rendering.

Check your three to five Pile A URLs in this order:

  1. Inspect the URL in Search Console. Is it indexed? Does the rendered page include the main answer text?
  2. Check the page source and the answer block. Look for noindex, nosnippet, max-snippet, or data-nosnippet instructions that limit what Google can show.
  3. Run PageSpeed Insights on mobile. If the answer arrives late or only after interaction, check the rendered result before assuming Google sees what you see.

I have seen copy get rewritten while a technical block sat untouched. That is expensive optimism. If a fix needs a developer, isolate the exact problem rather than sending a vague request to “optimize for AI.” This technical playbook for making a site readable to AI search is useful when the blocker is beyond a simple content edit.

What gets cut: design changes and schema experiments while the answer itself is missing from the rendered page. Check fetch and rendering before you edit the words. A cleanly rendered page can compete for a citation; it is not promised one.

4. Week 4: Rewrite two pages answer-first

Cost: 4 hours. Buys: two pages with answers Google can lift without assembling them from scattered paragraphs. Pick your top two Pile A pages. Do nothing else. For each page, turn the relevant H2 into the buyer’s exact question and put a direct answer immediately below it, before the backstory. Aim for roughly 40–55 words; that answer-first placement is a recurring recommendation in AI Overview tests.

The mechanism matters more than the word count. A complete answer should carry its number, scope, and condition together. If the first paragraph says “prices vary” and the useful range appears 400 words later, a reader has to hunt for it. So does any system trying to extract a short answer.

Say the H2 is “How much does garage door spring replacement cost in Austin?” The next paragraph should give the price range, say what moves it, and explain when the higher end applies. Then support it with a small table, a few bullets on what is included, and a line on timing if the page has that information. Do not invent a range to fill the format; use what the business can substantiate.

Give each page roughly 90 minutes for the rewrite and use the remaining hour to check both pages on mobile and confirm the answers appear where you put them. Leave useful sections of the existing page alone. What gets cut: new posts, an eight-question FAQ bolted onto each page, and every Pile C rewrite you thought of while doing this one.

If someone cannot copy the answer paragraph and understand it without the rest of the page, edit that paragraph again.

5. Week 5: Put proof beside the answers

Cost: 4 hours. Buys: a reason to cite your page rather than a generic answer. Return to the same two pages. Make the claims a buyer might repeat traceable: put a number beside its named source, link the primary document where you have one, and add a first-party fact the business can support. That might be a price range from your records, a service timeline, or the area you cover. Do not manufacture an impressive-looking count. No citation is worth making up the underlying work.

Then make ownership visible. Add a named author where appropriate, check that published and updated dates are accurate, and make any Article JSON-LD consistent with the headline, author, and dates readers can see. Schema is not a substitute for evidence on the page. It is particularly unhelpful when it disagrees with the page.

Budget one hour per page for sources and first-party details, 30 minutes per page for the byline and dates, and 30 minutes per page to check the markup in Rich Results Test. If a page has no defensible number, use the time to clarify the conditions behind its answer instead of forcing one in.

What gets cut: stock testimonials, badge walls, and generic claims that you “prioritize quality.” Reddit can supply anecdotes. A business page needs to say something useful that comes from the business. One supported first-party fact beats three paragraphs of reassurance.

6. Week 6: Run the same checks and decide what moved

Cost: 4 hours. Buys: a read on the two pages you changed. Re-run the same ten searches from Week 1 in the same logged-out browser. Record the same three fields: whether an Overview appears, whom it cites, and whether your site is among them. Then compare Search Console impressions and average position for your two rewritten pages with the four weeks before the edits.

Keep the interpretation narrow. Did a page gain a citation where it had none? Did its position change? Are the answer and its proof visible in the indexed, rendered page? An Overview can vary between checks, so a single search is a signal to record, not a victory lap. Citations also need time for pages to be recrawled and reconsidered. Checking on day three and rewriting again is not a plan.

Do not mistake thin clicks for proof that the page failed. The 300-query check found 1.7% CTR for pages at positions 7–10 under an Overview and 0.7% for positions 4–6. That is an observation from that check, not a rule for your site. Read your own position, impressions, and citation log together.

Use the final hour on the clearest remaining problem. If the answer appears cleanly but your page is not cited, revisit its proof before redesigning it. If the answer is absent from the rendered page, return to Week 3. If the question never triggers an Overview, stop treating it as an AI Overview target.

A new citation is progress. No citation is a result to diagnose, not a reason to buy software on Friday.

What I deliberately cut

I left out llms.txt, a new publishing cadence, an eight-question FAQ on every page, and video. Each costs time you do not have. None fixes a top-ranking page whose answer Google cannot readily find or read. Six thin new pages do not solve the problem on the page already sitting at position 8.

I used to tell small clients to publish weekly to stay fresh. For this constraint, I would rather spend the hours on two Pile A pages with clear answers and supportable details than on ten new posts without proof. The cuts protect the work that might earn the citation.

When the four-hour constraint lifts

Keep the manual check first: the same ten questions, once a month. Then apply the method to the next three Pile A pages and the follow-up questions attached to the Overviews you are targeting. Pay for software when checking your question set by hand costs more than the subscription; that becomes a more plausible trade once you are tracking dozens of queries or multiple locations. Before buying, look at which capabilities help you get featured rather than merely monitored.

If you would rather hand off the rewrites, crawl fixes, and repeat checks, groas runs search work as continuous execution with a human owner accountable for the result. Either way, the first 24 hours tell you which pages deserve more work. Spend the next dollar on those pages, not on a dashboard that tells you they exist.